atman-marketplace

Personal Claude Code marketplace collecting reusable plugins (skills, agents, hooks, MCP) for Windows and WSL development environments.

8 plugins 62 skills 4 sub-agents 3 MCP servers

Getting started

  1. Register this marketplace in Claude Code (once).

    /plugin marketplace add atman-33/atman-marketplace
  2. Install the plugins you need. Choose user scope in the /plugin dialog to make a plugin available across all projects.

    /plugin install <plugin>@atman-marketplace
  3. Follow the plugin's Setup section below — some plugins need extra commands or environment variables.

Skill index

Every skill across all plugins. Click a skill for its full behavior and output details; 📄 marks skills that produce a file or report (hover for where it goes).

PluginSkillTypeWhat it doesOutput
engineeringcapture-ruleautoQualifies an insight as reusable and non-obvious, routes it to exactly one home (target repo .claude/rules, harness .claude/rules-ex, harness .claude/rules, or auto-memory), dedups against existing rules on the same topic, and writes a focused kebab-case rule file with the correct paths: frontmatter for that home.📄
engineeringcommit-changesautoAssesses the working tree with git status/diff, stages everything if nothing is staged, drafts a Conventional Commits message matched to the diff, and commits.
engineeringcreate-adrautoFinds the target repo's existing ADR convention (or creates docs/adr/ starting at 0001), gathers the decision, context, alternatives, and consequences from the conversation, drafts a numbered ADR, and writes it after user approval.📄
engineeringcreate-feature-branchautoConfirms the target repository and the feature/branch name, verifies it is actually in that repo (git rev-parse --show-toplevel), updates the local main branch from the remote, then creates and switches to a feature/<slug> branch (following the repo's own prefix convention if different).
engineeringcreate-manual-test-guideautoDetermines the change under test (named feature/branch/PR or the current diff), reads the changed files to derive real test scenarios (happy path, edge cases, error handling, regression checks) and real launch/setup commands, then generates one self-contained interactive HTML guide with a test-flow SVG diagram, per-scenario pass/fail/skip toggles persisted to localStorage, and a Copy-results-as-Markdown export button.📄
engineeringcreate-onboarding-guideautoExplores the repository (layout, architecture traced from real imports, 2-3 key flows traced end-to-end with file:line hops, conventions and gotchas) and generates one self-contained HTML onboarding tour with an SVG architecture diagram, collapsible directory map, flow walkthroughs, a recommended reading order, and real getting-started commands.📄
engineeringcreate-pull-requestautoResolves the target repository from the active project context, runs bundled scripts (analyze_changes.mjs, generate_pr_body.mjs from a template matched to branch intent, quality_checks.mjs) in the target repo, then presents a localized PR title/body draft for approval before creating or updating the GitHub PR with gh.📄
engineeringcreate-review-guideautoDetermines the review scope (named range, dirty tree, or feature branch vs merge-base), reads every changed file in full and traces callers, then generates one self-contained HTML review guide with an overview, an SVG architecture diagram highlighting changed modules, a file/responsibility map, a collapsible annotated change walkthrough in recommended reading order, and a reviewer checklist.📄
engineeringdevelop-small-featureexplicitCarries a small, well-scoped change through the full lifecycle by chaining other components: create-feature-branch (if on the default branch), the tdd skill delegated to the implementer agent, static checks via the test-runner agent, an explicit stop for manual user verification of the uncommitted work, then commit-changes and create-pull-request.
engineeringinstall-recommended-skillsexplicitChecks gh auth, then runs gh skill install for six mattpocock/skills engineering skills (improve-codebase-architecture, tdd, to-issues, to-prd, codebase-design, grill-with-docs), continuing past individual failures and reporting a summary.
engineeringinvestigate-bug-reportautoDiagnoses a reported bug without changing code: pins the report down to one given-X-does-Y-should-do-Z statement, reproduces the failure (or statically traces the code path when reproduction is infeasible), isolates the root cause to a specific file:line with git blame corroboration, classifies every caller for blast radius, and proposes 1-3 fix candidates plus the regression test that would have caught it.
engineeringprepare-releaseautoCreates a release/v<VERSION> branch from the up-to-date default branch, bumps the version in every version-bearing file (regenerating lockfiles), shows the commit history since the last tag and curates CHANGELOG.md with the user (Keep a Changelog format), then commits (chore(release): prepare v<VERSION>) and opens a PR to the default branch with the changelog as the body.📄
engineeringset-openspec-pathexplicitReads .claude/project-context.json, presents the registered projects as an AskUserQuestion menu (plus a Clear option), and rewrites only the openspecPath value to <chosen project>/openspec, preserving every other field.📄
engineeringsetup-allexplicitRuns all four engineering setup phases in sequence, continuing past failures: installs the recommended mattpocock/skills set via gh, installs the OpenSpec CLI globally and runs openspec init --tools claude, scaffolds .claude/project-context.json if missing, and scaffolds the rules-ex files from the setup-rules-ex templates.📄
engineeringsetup-openspecexplicitVerifies Node.js >= 20.19.0, installs @fission-ai/openspec globally if absent, runs openspec init --tools claude in the current project, and reports the newly available /opsx:* commands.📄
engineeringsetup-project-contextexplicitShows the existing .claude/project-context.json if present (and stops), otherwise creates it from a template with roleBasedDelegation, openspecPath, postToolFormatCommands, and projects placeholders, then explains how each field drives the plugin's SessionStart/PreToolUse/PostToolUse hooks and that changes take effect at the next session start.📄
engineeringsetup-rules-exexplicitScaffolds the two files the inject-extended-rules hook needs, copying the canonical templates from the skill's assets/templates/ and never overwriting existing files: a path-scoped authoring-guide rule and the rules-ex README.📄
obsidiandefuddleautoFetches a web page with the Defuddle CLI (defuddle parse <url> --md) instead of WebFetch, stripping navigation, ads, and clutter to return clean Markdown (or specific metadata via -p), saving tokens.
obsidianjson-canvasautoCreates and edits .canvas files per the JSON Canvas 1.0 spec: generates unique 16-hex-char node/edge IDs, positions nodes without overlap, connects them with validated fromNode/toNode edges, and re-validates JSON and reference integrity after every change.📄
obsidiankb-indexautoUpdates PARA Knowledge Base indexes using a 3-tier strategy: by default a smart diff that compares each _index.md against the files actually on disk and processes only additions/removals; --full rebuilds every index from the template; --category limits scope; --dry-run previews.📄
obsidiankb-ingestautoProcesses new documents (from Inbox/ or a given path) into the PARA vault: reads and classifies each document to a PARA category using index files for context, plans a subdirectory-first destination and canonical filename before moving, generates frontmatter, creates wikilinks and backlinks, and updates the affected indexes and activity log.📄
obsidiankb-initautoSets up the Knowledge Base layer on an existing PARA-structured vault: detects the vault root via .obsidian/, generates an _index.md in each PARA category from the template, creates the top-level index and append-only activity log, appends KB rules to AGENTS.md, and detects whether the obsidian CLI is available.📄
obsidiankb-lintautoAudits the vault's health read-only (or auto-fixes safe issues with --fix): index drift against files on disk, orphan documents with no links/index entries/tags, tag consistency (#proj/ and #type/ conventions), and broken wikilinks — via the obsidian CLI when available, Grep otherwise.
obsidiankb-queryautoAnswers questions about vault content by picking the fastest search route (direct folder access, tag cross-collection, backlink traversal, or index+log browse) rather than always starting from the top-level index, synthesizes an answer across documents with wikilink citations, and can optionally save the insight back as a vault note.
obsidianobsidian-basesautoCreates and edits Obsidian .base files (YAML database views over notes): defines filters, computed formula properties, property display config, summaries, and one or more table/cards/list/map views, then validates YAML syntax, quoting rules, and formula references.📄
obsidianobsidian-cliautoReference for driving a running Obsidian instance through the obsidian CLI: parameter/flag syntax, file targeting by wikilink-style name or exact path, vault targeting, and common commands for reading/creating/appending notes, search, daily notes, properties, tasks, tags, and backlinks — plus plugin/theme development commands.
obsidianobsidian-markdownautoSyntax reference for authoring valid Obsidian Flavored Markdown: wikilinks (heading/block variants), embeds (![[...]] for notes, images, PDFs), callouts, frontmatter properties, and when to prefer wikilinks over standard Markdown links — with deeper references in bundled PROPERTIES/EMBEDS/CALLOUTS files.
obsidianzenn-blog-writingautoJapanese-language guideline skill for writing, editing, and reviewing Zenn tech-blog articles: the draft-to-publish workflow (published: false, npm run textlint, preview, publish), required frontmatter and topics naming rules, and prose-quality rules that emphasize technical accuracy and avoiding AI-sounding phrasing.
obsidianzenn-markdownautoSyntax reference for Zenn-specific Markdown: message blocks, accordions (:::details), code-block filename/diff extensions, embedded content (link cards, videos, SNS), KaTeX math, mermaid diagrams, and image size/caption controls.
productivitycreate-claude-commandautoReference guide for authoring Claude Code slash commands in a project's .claude/commands/: the scaffold template, $ARGUMENTS/$1 handling, @file and !command injection, naming conventions, and when to use a command vs agent vs skill.📄
productivitycreate-claude-mdautoInspects the project (build/test/lint config, README, git log conventions) and writes a concise CLAUDE.md that includes only what Claude cannot infer (commands, non-default style rules, workflow conventions, gotchas), routing domain-specific rules to .claude/rules/<domain>.md stubs and keeping the file under ~50 lines where possible.📄
productivitycreate-readmeautoInvestigates the project first (audience/README type, identity, run commands, features, assets, git remote), then writes a README ordered around the dominant intent (consumption, library, or contribution focused) using a prioritized section menu, GFM formatting, verified badges, and GitHub admonitions — never inventing commands or badges it cannot verify.📄
productivitycreate-release-notesautoCollects commits between the last tag and HEAD (or a given range), enriches thin commits from merged PRs via gh, filters out internal-only changes, and writes user-facing notes ordered by impact (Breaking/New/Improved/Fixed) with each item stating the user benefit rather than the implementation.📄
productivitycreate-work-logautoScans the repositories configured in the skill's config.json for git activity in the period (commits on all branches by the configured author, plus uncommitted work), groups it by repo and Conventional Commit theme, asks the user for anything git can't show (or notes auto-generated when unattended), and saves a Markdown log.📄
productivitygenerate-html-reportautoPicks one of five report modes (spec/plan, code-review explainer, design prototype, research/status report, or throwaway editing UI), gathers the real content first (source, git, MCPs, web), and writes one self-contained HTML file with inline CSS/JS/SVG — used in place of any Markdown document that would cross ~100 lines or needs diagrams/interactivity.📄
productivityhandoff-goexplicitWrites a handoff document summarizing the current conversation (with a suggested-skills section, secrets redacted) to the OS temp directory, then runs the bundled dispatch.mjs which launches a fresh claude session that reads it — splitting a multiplexer pane (herdr/zellij), opening a new terminal window, or printing a paste-ready command as fallbacks.📄
productivityherdrautoTeaches control of the herdr multiplexer from inside it (requires HERDR_ENV=1): listing workspaces/tabs/panes, reading other panes' output, splitting panes and running commands in them, waiting on output or another agent's status, and sending text/keys — all via herdr CLI commands over a local unix socket.
productivityinstall-recommended-skillsexplicitChecks gh auth, then installs three mattpocock/skills productivity skills (grill-me, handoff, writing-great-skills) at user scope via gh skill install, continuing past individual failures and reporting a summary.
productivityinstall-skillexplicitParses a GitHub skills-repo tree URL (or an OWNER/REPO + path reference) into a repo and skill path, then installs it with gh skill install <repo> <skill-path> --agent claude-code and reports the result.
productivitylaunch-teamexplicitActs as conductor for a multi-agent team: settles 2-4 roles (one orchestrator), writes a self-contained brief and a team config JSON to the OS temp dir, then runs the bundled launcher.mjs which opens one named multiplexer pane per agent (herdr or zellij), writes the shared roster, and delivers the brief to the orchestrator.
productivitymanage-desktop-routinesexplicitThree branches for Claude Code Desktop local routines: create (collects taskId/description/self-contained prompt/schedule, calls create_scheduled_task, hands the user a post-creation checklist for folder/model/permissions), backup (dumps all tasks with prompts to a JSON file), and restore (recreates tasks missing on this machine from the backup, never overwriting existing ones).📄
productivitypost-to-slackexplicitResolves a channel name to its Slack ID via slack_search_channels (and a mention name to <@USERID> via slack_search_users), then posts with slack_send_message (supporting thread replies) and returns the permalink.
productivitysetup-herdrexplicitChecks for herdr in both Windows and WSL (PATH, known install dirs, running process), installs it via the official install scripts only after user confirmation, installs the Claude Code/OpenCode integrations, and edits shell profiles (PowerShell, bash, zsh) to auto-launch herdr — removing any zellij autostart block first.
productivitysetup-zellijexplicitChecks for zellij in PowerShell and WSL, installs it (winget / apt / cargo) after user confirmation, then edits the four shell profiles to auto-launch zellij on startup, removing any herdr autostart block installed by setup-herdr first.
productivitysidekick-goautoSpawns a persistent helper agent in an adjacent multiplexer pane (herdr or zellij): writes a self-contained brief to the OS temp dir, runs the bundled dispatch.mjs to open the sidekick's pane and wire two-way messaging over the shared bus, then iterates with it across rounds (replies arrive in the caller's own prompt) until a closing message ends the exchange.
productivitywsl-vscode-doctorexplicitDiagnoses why code .📄
scrumaudit-epic-consistencyautoCross-checks an Epic's backlog against its code: refreshes the snapshot and repo sync if stale, uses .pm/repo/branch-diff.json to scope the investigation to files the Epic actually touched, audits every in-scope item for drift, gaps (missing acceptance.md), and clarity with file:line evidence from the dedicated read-only clone, then applies only user-approved fixes (acceptance.md edits directly, monday edits via monday-api-mcp).📄
scruminvestigate-spikeautoIdentifies the spike item on the monday board, loads its question from the item plus saved updates and acceptance.md, syncs the Epic's dedicated repo clone (sync-repo.mjs), investigates read-only with file:line evidence until there is enough to decide, then writes a findings report and offers to post the conclusion back to the monday item.📄
scrumload-epic-contextautoPrimes a session with an Epic's state at minimal token cost: refreshes the backlog snapshot (save-all.mjs) and repo sync (sync-repo.mjs) when older than 60 minutes, then reads only the aggregates (summary.md, progress-history.json, repo-state.json, branch-diff.json, tail of commits.jsonl) — never individual item files — and briefs the user before offering to (re)write the summary.📄
scrummanage-drive-docsautoFinds, reads, edits, and reorganizes development docs in Google Drive by working directly on the locally synced Google Drive for desktop folder with standard file tools scoped to the configured driveDocsRootPath, falling back to the bundled drive-fs.mjs script for rename/move/delete/list operations and for non-ASCII WSL paths the standard tools cannot reach.
scrummanage-monday-backlogautoReads and updates the monday.com backlog primarily through the plugin's dependency-free CLI scripts (list-items, get-item, update-item-status, save-item/doc/updates/all, init-task, set-link, check-stale-links) which need only MONDAY_TOKEN and Node 18+, resolving the board id and Epic folders from .claude/scrum-context.json; the monday-api-mcp server is the optional path for operations beyond the scripts' surface.📄
scrumprepare-sprint-reviewautoRefreshes the Epic snapshot with save-all.mjs, classifies items in the sprint window (done / progressed / carried over), gathers demo material per Done item from its pbi folder's acceptance.md and evidence/ (marking items with no recorded evidence), and generates a self-contained HTML review document with metrics, a demo agenda with time-boxes, per-item cards, and discussion prompts.📄
scrumrefine-backlogautoAudits every non-Done item (or a narrowed scope) against four readiness checks — acceptance criteria present, estimate present, size within threshold, outcome-stating title — proposes concrete fixes (drafted Given/When/Then criteria, concrete splits, rewritten titles; never invented estimates), and applies only user-approved changes: criteria to the item's acceptance.md, board edits via the monday-api-mcp server or a paste-ready list.
scrumreport-epic-specexplicitReconciles what the product does today vs what is planned by reading both sides: board items by status from .pm/backlog/ plus prd/prd.md, and the code via branch-diff.json, recent commits, and targeted reads in the dedicated clone — flagging Done items with no implementation trace and implemented behavior no item records — then writes a stakeholder-readable HTML report.📄
scrumreport-pbl-progressexplicitRuns the bundled generate-progress-report.mjs once against the Epic's Drive snapshot, which computes totals, status mix, and a completion forecast (assumed velocity until 2+ days of history exist, then measured) and renders the HTML report itself; the skill then resolves the report's Drive link and sharing status via the Drive MCP connector and reports the KPIs — never re-deriving aggregates from item files.📄
scrumsetup-scrum-contextautoCreates or updates the project's .claude/scrum-context.json without hand-editing: shows current values, extracts the board id from a pasted monday URL, sanity-checks that new mondayEpics folders fall under driveDocsRootPath, then pipes a JSON patch to the bundled init-scrum-context.mjs which merges onto the existing config and ensures the file is git-ignored.📄
scrumsnapshot-pbl-to-driveexplicitOne-shot wrapper that runs save-all.mjs once for the named Epic (board id and Drive folder auto-resolve from scrum-context) and reports the script's JSON summary — deliberately never opening individual snapshot files afterward.📄
scrumwrite-retrospectiveautoFacilitates then records a KPT retrospective: prepares sprint facts from the board (done/carry-over, plus a follow-up verdict on every previous Try action), walks the user through Keep/Problem/Try one pass at a time with data-grounded candidate observations, turns Try items into owned, dated actions, saves the record, and offers to put actions on the monday board via monday-api-mcp.📄
stack-cloudflarecloudflare-static-assetsautoGuides configuring Cloudflare Static Assets for React Router apps on Workers — adding the assets binding to wrangler.jsonc and creating a utility module that fetches public/ files uniformly across dev (Vite fetch), production (ASSETS binding), and build time (filesystem) — fixing the dev-works/prod-404 mismatch.
stack-dnd-kitdnd-kit-implementationautoProvides the implementation pattern for combining useSortable and useDroppable on the same container in dnd-kit (Kanban-style UIs): detect what is being dragged via useDndContext and conditionally switch refs/listeners so containers act as drop targets when items drag and as sortables when containers drag.
stack-opencodeopencode-log-investigationautoDiagnoses OpenCode runtime failures by opening the newest log under ~/.local/share/opencode/log/ (or the Windows equivalent) and searching for ERROR lines, with targeted workflows for plugin-loading issues (service=plugin), MCP server failures (service=mcp), session/tool errors, and the gotcha that plugin stdout/stderr never reaches the log file.
stack-react-routerreact-router-v7-appautoPrescribes React Router v7 conventions: file-based routing in app/routes/ (folder routes, route groups, dot-notation nesting, excluding test files from route discovery), loader/action data patterns with typed Route.* props, arrow-function route components, ErrorBoundary usage, and meta/headers exports.

engineering v0.19.0

Engineering utilities and helpers for software development tasks.

Setup

  1. Install the plugin (project or user scope).

    /plugin install engineering@atman-marketplace
  2. Prerequisites: Node.js 20+, gh (authenticated) on PATH, and uv/uvx on PATH for the serena MCP server.

  3. Run the all-in-one setup: installs the recommended mattpocock/skills set, installs the OpenSpec CLI and runs openspec init, and scaffolds .claude/project-context.json plus the rules-ex files. Individual steps are also available as setup-openspec / setup-project-context / setup-rules-ex.

    /setup-all
  4. Confirm the serena and context7 MCP servers are connected.

    /mcp

Typical workflow

After one-time setup, the skills map onto the everyday feature-development loop; the sub-agents handle delegated execution inside it.

  1. One-time setup

    Or run setup-openspec / setup-project-context / setup-rules-ex selectively.

  2. Start a change
  3. Implement

    develop-small-feature chains the whole loop (branch, TDD via sub-agents, checks, commit, PR) for small changes; investigate-bug-report diagnoses defects without changing code.

  4. Commit and open a PR
  5. Verify and review

    Generate self-contained HTML guides for manual testing, code review, and onboarding.

  6. Capture knowledge as you go

    Persist non-obvious insights and design decisions at the moment of discovery.

  7. Ship

Skills 17

capture-rule auto-invocable

Capture a reusable, non-obvious project insight into the rules home where it will auto-inject in future sessions. Use when investigation or implementation surfaces knowledge worth persisting (a gotcha, a convention, the "why" behind a decision), or when the user asks to remember a project learning.

Behavior

Qualifies an insight as reusable and non-obvious, routes it to exactly one home (target repo .claude/rules, harness .claude/rules-ex, harness .claude/rules, or auto-memory), dedups against existing rules on the same topic, and writes a focused kebab-case rule file with the correct paths: frontmatter for that home.

📄 Output

A rule file: <target-repo>/.claude/rules/<slug>.md, harness .claude/rules-ex/<slug>.md, or harness .claude/rules/<slug>.md depending on routing (auto-memory insights are not written as rules).

Arguments: The insight to capture, and which repository it applies to.

commit-changes auto-invocable

Commit the current changes with a Conventional Commits message. Use when the user asks to commit, or when a workflow (e.g. develop-small-feature) reaches its commit step after implementation is verified.

Behavior

Assesses the working tree with git status/diff, stages everything if nothing is staged, drafts a Conventional Commits message matched to the diff, and commits. Stops and reports if there is nothing to commit or the changes look like multiple unrelated commits.

create-adr auto-invocable

Record an architecture/design decision as an ADR (Architecture Decision Record) in the target repository's docs/adr/. Use when the user wants to document a design decision, its alternatives and consequences, or says "record this decision" / "write an ADR".

Behavior

Finds the target repo's existing ADR convention (or creates docs/adr/ starting at 0001), gathers the decision, context, alternatives, and consequences from the conversation, drafts a numbered ADR, and writes it after user approval. If the new ADR supersedes an old one it also flips the old ADR's Status.

📄 Output

Markdown ADR written to <target-repo>/docs/adr/NNNN-<kebab-slug>.md (or the repo's existing ADR directory convention).

create-feature-branch auto-invocable

Creates a new feature branch from main inside the intended repository after confirming the branch target and repository context. Use when the user wants to start a new feature branch, asks to create a branch for upcoming work, mentions a branch name such as feature/<name>, or another workflow has already resolved the target repository.

Behavior

Confirms the target repository and the feature/branch name, verifies it is actually in that repo (git rev-parse --show-toplevel), updates the local main branch from the remote, then creates and switches to a feature/<slug> branch (following the repo's own prefix convention if different).

Arguments: What feature or branch name should this use?

create-manual-test-guide auto-invocable

Generate a self-contained HTML manual-testing guide for an implemented change — environment setup, a visual test-flow diagram, step-by-step scenarios with expected results, and interactive pass/fail checkboxes with a Markdown results export. Use when the user wants to verify an implementation by hand, asks "how do I test this", or wants a verification document for a feature/branch/PR.

Behavior

Determines the change under test (named feature/branch/PR or the current diff), reads the changed files to derive real test scenarios (happy path, edge cases, error handling, regression checks) and real launch/setup commands, then generates one self-contained interactive HTML guide with a test-flow SVG diagram, per-scenario pass/fail/skip toggles persisted to localStorage, and a Copy-results-as-Markdown export button.

📄 Output

Self-contained HTML named <repo>-<branch-or-feature-slug>-manual-test-<yyyyMMdd-HHmm>.html written to the outputDir configured in the plugin-level or skill-level config.json (copied from config.example.json on first run), then auto-opened in the browser when openAfterGenerate is true (default).

create-onboarding-guide auto-invocable

Generate a self-contained HTML onboarding tour of a repository — architecture diagram, directory map, key flow walkthroughs, and a recommended reading order — for new team members. Use when the user wants onboarding material, asks for a codebase overview/architecture tour, or wants to explain a repo to someone new.

Behavior

Explores the repository (layout, architecture traced from real imports, 2-3 key flows traced end-to-end with file:line hops, conventions and gotchas) and generates one self-contained HTML onboarding tour with an SVG architecture diagram, collapsible directory map, flow walkthroughs, a recommended reading order, and real getting-started commands.

📄 Output

Self-contained HTML named <repo>-onboarding-<yyyyMMdd>.html written to the outputDir configured in the plugin-level or skill-level config.json, auto-opened when openAfterGenerate is true (default).

create-pull-request auto-invocable requires: Requires Node 18.3+, git, and GitHub CLI. PowerShell examples assume Windows.

Analyzes git changes, drafts localized PR titles and bodies, and assists with creating or updating GitHub pull requests for the active agent-harness project repository. Use when working from agent-harness and the user wants to create a PR, review branch changes, draft or update a PR description, or check whether a branch is ready for review.

Behavior

Resolves the target repository from the active project context, runs bundled scripts (analyze_changes.mjs, generate_pr_body.mjs from a template matched to branch intent, quality_checks.mjs) in the target repo, then presents a localized PR title/body draft for approval before creating or updating the GitHub PR with gh.

📄 Output

A GitHub pull request (created or updated via gh); intermediate analysis and PR-body files are kept only under the skill's own .tmp/ directory, never in the target repo.

create-review-guide auto-invocable

Generate a self-contained HTML code-review guide for the current changes — big-picture overview, architecture diagram of affected modules, a file/class responsibility map, and an annotated change walkthrough — so a reviewer can grasp the whole implementation before diving into the diff. Use when the user wants review material for a branch/PR, asks to "explain the changes for review", or wants a visual summary of what was implemented where.

Behavior

Determines the review scope (named range, dirty tree, or feature branch vs merge-base), reads every changed file in full and traces callers, then generates one self-contained HTML review guide with an overview, an SVG architecture diagram highlighting changed modules, a file/responsibility map, a collapsible annotated change walkthrough in recommended reading order, and a reviewer checklist.

📄 Output

Self-contained HTML named <repo>-<branch-slug>-review-<yyyyMMdd-HHmm>.html written to the outputDir configured in the plugin-level or skill-level config.json, auto-opened when openAfterGenerate is true (default).

develop-small-feature explicit-only

Implement a small, well-scoped feature or fix end-to-end from spec to PR — reuse or create a feature branch, implement test-first, run static checks, get user verification, commit, then open a PR to main.

Behavior

Carries a small, well-scoped change through the full lifecycle by chaining other components: create-feature-branch (if on the default branch), the tdd skill delegated to the implementer agent, static checks via the test-runner agent, an explicit stop for manual user verification of the uncommitted work, then commit-changes and create-pull-request. Explicit-invocation only.

Arguments: The feature or fix spec to implement (optional — falls back to conversation context).

investigate-bug-report auto-invocable

Investigate a reported bug to its root cause — reproduce, trace the code path, identify the defect with evidence, assess blast radius, and propose fix candidates without changing code. Use when the user reports or pastes a bug/defect/unexpected behavior and wants it diagnosed, or asks "why is X happening".

Behavior

Diagnoses a reported bug without changing code: pins the report down to one given-X-does-Y-should-do-Z statement, reproduces the failure (or statically traces the code path when reproduction is infeasible), isolates the root cause to a specific file:line with git blame corroboration, classifies every caller for blast radius, and proposes 1-3 fix candidates plus the regression test that would have caught it.

prepare-release auto-invocable

Prepares a release by creating a release branch, bumping the version, guiding manual changelog curation from git history, and opening a PR to main. Use when the user wants to cut a release, prepare a release PR, ship a version, mentions "release prep", "prepare release", "cut release", or wants to manually write a changelog instead of auto-generating one.

Behavior

Creates a release/v<VERSION> branch from the up-to-date default branch, bumps the version in every version-bearing file (regenerating lockfiles), shows the commit history since the last tag and curates CHANGELOG.md with the user (Keep a Changelog format), then commits (chore(release): prepare v<VERSION>) and opens a PR to the default branch with the changelog as the body.

📄 Output

Version bumps and a curated CHANGELOG.md entry committed to a release/v<VERSION> branch, plus a GitHub PR to the default branch.

set-openspec-path explicit-only

Switch the openspecPath in .claude/project-context.json by picking one of the registered projects from a menu, instead of hand-editing the absolute path.

Behavior

Reads .claude/project-context.json, presents the registered projects as an AskUserQuestion menu (plus a Clear option), and rewrites only the openspecPath value to <chosen project>/openspec, preserving every other field. Reminds the user the change takes effect at the next session start. Explicit-invocation only.

📄 Output

Updated openspecPath value in the current project's .claude/project-context.json.

setup-all explicit-only

Run all engineering plugin setup steps in sequence — install recommended skills, set up OpenSpec, scaffold project-context.json, and set up rules-ex infrastructure.

Behavior

Runs all four engineering setup phases in sequence, continuing past failures: installs the recommended mattpocock/skills set via gh, installs the OpenSpec CLI globally and runs openspec init --tools claude, scaffolds .claude/project-context.json if missing, and scaffolds the rules-ex files from the setup-rules-ex templates. Ends with a per-phase status table. Explicit-invocation only.

📄 Output

OpenSpec scaffolding in the project, .claude/project-context.json, .claude/rules/rules-ex-authoring.md, and .claude/rules-ex/README.md (each only created if missing).

setup-openspec explicit-only

Install OpenSpec CLI (if not already installed) and run openspec init --tools claude to add spec-driven development skills to this project.

Behavior

Verifies Node.js >= 20.19.0, installs @fission-ai/openspec globally if absent, runs openspec init --tools claude in the current project, and reports the newly available /opsx:* commands. Explicit-invocation only.

📄 Output

OpenSpec scaffolding created in the current project by openspec init --tools claude.

setup-project-context explicit-only

Scaffold or show the .claude/project-context.json file that the engineering hooks read to inject project context, target-repo guidance, and optional post-edit format commands.

Behavior

Shows the existing .claude/project-context.json if present (and stops), otherwise creates it from a template with roleBasedDelegation, openspecPath, postToolFormatCommands, and projects placeholders, then explains how each field drives the plugin's SessionStart/PreToolUse/PostToolUse hooks and that changes take effect at the next session start. Explicit-invocation only.

📄 Output

.claude/project-context.json created in the current project root (if it did not already exist).

setup-rules-ex explicit-only

Scaffold the rules-ex extended-rules infrastructure (.claude/rules/rules-ex-authoring.md and .claude/rules-ex/README.md) required by the engineering plugin's inject-extended-rules hook.

Behavior

Scaffolds the two files the inject-extended-rules hook needs, copying the canonical templates from the skill's assets/templates/ and never overwriting existing files: a path-scoped authoring-guide rule and the rules-ex README. Ends with a created/already-existed status table. Explicit-invocation only.

📄 Output

.claude/rules/rules-ex-authoring.md and .claude/rules-ex/README.md in the current project (only created if missing).

Sub-agents 4

AgentModelWhen to useTools
code-exploresonnetUse for broad, read-only investigation — sweeping many files or tracing where symbols are defined, called, or configured (via serena's symbol-aware tools) — when you need the conclusion as file:line references, not edits.Read, Grep, Glob, mcp__serena__initial_instructions, mcp__serena__activate_project, mcp__serena__find_symbol, mcp__serena__find_referencing_symbols, mcp__serena__find_declaration, mcp__serena__find_implementations, mcp__serena__get_symbols_overview, mcp__serena__search_for_pattern
heavy-implementersonnetUse for large or uncertain multi-file changes that need their own build/test/debug iteration loop in an isolated context; it can invoke skills (tdd, verify) and delegate sub-tasks to test-runner or code-explore.Read, Grep, Glob, Edit, Write, Bash, Skill, Agent, mcp__serena, mcp__context7
implementersonnetUse for changes whose specification is already settled — mechanical, well-scoped edits (or a batch of small related ones) applied cleanly with serena's precise symbol-editing tools, no re-litigating the approach.Read, Grep, Glob, Edit, Write, mcp__serena
test-runnerhaikuUse to run a known test/build/lint command and fold only a concise pass/fail verdict (with key failure lines) back into the main session, keeping verbose logs out of the main context; runs on haiku.Bash, Read, Grep, Glob

Hooks 3

SessionStart

node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/inject-project-context.mjs"

What it does

inject-project-context.mjs reads .claude/project-context.json from the project root and injects a <project-context> block (registered project paths + resolved openspec docs folder) into Claude's context, plus the role-based delegation criteria when roleBasedDelegation is true. Silent when the config file is absent.

PreToolUse matcher: Read|Edit|Write

node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/inject-target-rules.mjs"

node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/inject-extended-rules.mjs"

What it does

On Read/Edit/Write, inject-target-rules.mjs resolves the touched file against registered sibling projects in .claude/project-context.json and injects that repo's CLAUDE.md/AGENTS.md in full (once per session) plus any path-matching .claude/rules/*.md; inject-extended-rules.mjs does the same for the workspace's .claude/rules-ex/ rules (workspace-relative paths: globs, strict matching).

PostToolUse matcher: Edit|Write

node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/post-format-project.mjs"

What it does

On Edit/Write of a file under a registered target project outside the cwd, post-format-project.mjs runs the postToolFormatCommands from .claude/project-context.json (per-project entry overrides the top-level default) sequentially in that project's root, best-effort and non-blocking, reporting results in a systemMessage.

MCP servers 2

context7

node ${CLAUDE_PLUGIN_ROOT}/mcp/context7-mcp-launcher.mjs

Notes

Up-to-date library/framework documentation lookup (@upstash/context7-mcp) launched via a Node launcher that falls back to npx; no API key required, and a global npm install (npm i -g @upstash/context7-mcp) only speeds startup.

serena

node ${CLAUDE_PLUGIN_ROOT}/mcp/serena-mcp-launcher.mjs

Notes

Semantic code retrieval/editing toolkit (oraios/serena) launched via a cross-platform Node launcher that runs uvx with Python pinned to 3.11 (--native-tls on Windows, login shell on WSL); requires uv/uvx on PATH, and the first launch pulls Serena from git.

obsidian v0.2.0

Helpers for working with Obsidian notes and knowledge management.

Setup

  1. Install the plugin.

    /plugin install obsidian@atman-marketplace
  2. Per-skill CLI prerequisites: defuddle needs the Defuddle CLI; obsidian-cli needs the Obsidian CLI with Obsidian running.

    npm install -g defuddle
  3. For the kb-* skills, initialize the Knowledge Base layer on your PARA-structured vault first.

    /kb-init

Typical workflow

The kb-* family forms a pipeline over a PARA-structured vault; the other skills are standalone syntax references and tools.

  1. Initialize the KB layer once

    Creates the per-category indexes, top-level index, activity log, and KB rules in AGENTS.md.

  2. Ingest new documents

    Typically from the vault's Inbox/ — classifies, moves, links, and indexes each document.

  3. Use and maintain the KB

    Query for answers; lint and re-index periodically to keep the vault healthy.

Skills 12

defuddle auto-invocable

Extract clean markdown content from web pages using Defuddle CLI, removing clutter and navigation to save tokens. Use instead of WebFetch when the user provides a URL to read or analyze, for online documentation, articles, blog posts, or any standard web page. Do NOT use for URLs ending in .md — those are already markdown, use WebFetch directly.

Behavior

Fetches a web page with the Defuddle CLI (defuddle parse <url> --md) instead of WebFetch, stripping navigation, ads, and clutter to return clean Markdown (or specific metadata via -p), saving tokens. Not for URLs that are already .md.

json-canvas auto-invocable

Create and edit JSON Canvas files (.canvas) with nodes, edges, groups, and connections. Use when working with .canvas files, creating visual canvases, mind maps, flowcharts, or when the user mentions Canvas files in Obsidian.

Behavior

Creates and edits .canvas files per the JSON Canvas 1.0 spec: generates unique 16-hex-char node/edge IDs, positions nodes without overlap, connects them with validated fromNode/toNode edges, and re-validates JSON and reference integrity after every change.

📄 Output

.canvas files created or edited in the Obsidian vault.

kb-index auto-invocable

Update Knowledge Base indexes. Smart mode detects changes and updates only what's needed. Full rebuild available with --full flag. Use after adding/moving documents or when indexes feel stale.

Behavior

Updates PARA Knowledge Base indexes using a 3-tier strategy: by default a smart diff that compares each _index.md against the files actually on disk and processes only additions/removals; --full rebuilds every index from the template; --category limits scope; --dry-run previews.

📄 Output

Updated _index.md files per PARA category and the top-level index in the vault.

Arguments: [--full] [--category <name>] [--dry-run]

kb-ingest auto-invocable

Ingest documents into the PARA Knowledge Base. Classifies, moves, links, and indexes new documents from Inbox or a specified path. Use when new documents need processing into the vault.

Behavior

Processes new documents (from Inbox/ or a given path) into the PARA vault: reads and classifies each document to a PARA category using index files for context, plans a subdirectory-first destination and canonical filename before moving, generates frontmatter, creates wikilinks and backlinks, and updates the affected indexes and activity log. Inbox README.md files are treated as structural notes and excluded from batch ingest.

📄 Output

Documents moved/renamed into their PARA destination folders in the vault, with updated frontmatter, wikilinks, indexes, and log entries.

Arguments: [file-or-folder-path]

kb-init auto-invocable

Initialize PARA Knowledge Base in your Obsidian vault. Creates index files, log, and adds KB rules to AGENTS.md. Run this first before using other kb- skills.

Behavior

Sets up the Knowledge Base layer on an existing PARA-structured vault: detects the vault root via .obsidian/, generates an _index.md in each PARA category from the template, creates the top-level index and append-only activity log, appends KB rules to AGENTS.md, and detects whether the obsidian CLI is available.

📄 Output

_index.md in each PARA directory, '0. Common/index.md', '0. Common/log.md', and a KB-rules section appended to the vault's AGENTS.md.

Arguments: [--force]

kb-lint auto-invocable

Health check for the PARA Knowledge Base. Detects orphan documents, broken links, index drift, tag issues, and stale content. Run periodically or as part of weekly review.

Behavior

Audits the vault's health read-only (or auto-fixes safe issues with --fix): index drift against files on disk, orphan documents with no links/index entries/tags, tag consistency (#proj/ and #type/ conventions), and broken wikilinks — via the obsidian CLI when available, Grep otherwise.

Arguments: [--fix] [--category <name>]

kb-query auto-invocable

Query the PARA Knowledge Base to find information, synthesize answers from multiple documents, and optionally save insights. Use when asking questions about vault content or needing cross-document analysis.

Behavior

Answers questions about vault content by picking the fastest search route (direct folder access, tag cross-collection, backlink traversal, or index+log browse) rather than always starting from the top-level index, synthesizes an answer across documents with wikilink citations, and can optionally save the insight back as a vault note.

Arguments: <question>

obsidian-bases auto-invocable

Create and edit Obsidian Bases (.base files) with views, filters, formulas, and summaries. Use when working with .base files, creating database-like views of notes, or when the user mentions Bases, table views, card views, filters, or formulas in Obsidian.

Behavior

Creates and edits Obsidian .base files (YAML database views over notes): defines filters, computed formula properties, property display config, summaries, and one or more table/cards/list/map views, then validates YAML syntax, quoting rules, and formula references.

📄 Output

.base files created or edited in the Obsidian vault.

obsidian-cli auto-invocable

Interact with Obsidian vaults using the Obsidian CLI to read, create, search, and manage notes, tasks, properties, and more. Also supports plugin and theme development with commands to reload plugins, run JavaScript, capture errors, take screenshots, and inspect the DOM. Use when the user asks to interact with their Obsidian vault, manage notes, search vault content, perform vault operations from the command line, or develop and debug Obsidian plugins and themes.

Behavior

Reference for driving a running Obsidian instance through the obsidian CLI: parameter/flag syntax, file targeting by wikilink-style name or exact path, vault targeting, and common commands for reading/creating/appending notes, search, daily notes, properties, tasks, tags, and backlinks — plus plugin/theme development commands. Requires Obsidian to be open.

obsidian-markdown auto-invocable

Create and edit Obsidian Flavored Markdown with wikilinks, embeds, callouts, properties, and other Obsidian-specific syntax. Use when working with .md files in Obsidian, or when the user mentions wikilinks, callouts, frontmatter, tags, embeds, or Obsidian notes.

Behavior

Syntax reference for authoring valid Obsidian Flavored Markdown: wikilinks (heading/block variants), embeds (![[...]] for notes, images, PDFs), callouts, frontmatter properties, and when to prefer wikilinks over standard Markdown links — with deeper references in bundled PROPERTIES/EMBEDS/CALLOUTS files.

zenn-blog-writing auto-invocable

Zennの技術ブログ記事を執筆・編集・レビューするためのガイドラインを提供する。技術的正確性、読みやすさ、AIっぽい文章の排除を重視。Zennの記事(articles/*.md)を書く・直す・レビューするとき、または「Zenn」「技術ブログ」「記事執筆」に言及されたときに使用する。

Behavior

Japanese-language guideline skill for writing, editing, and reviewing Zenn tech-blog articles: the draft-to-publish workflow (published: false, npm run textlint, preview, publish), required frontmatter and topics naming rules, and prose-quality rules that emphasize technical accuracy and avoiding AI-sounding phrasing.

zenn-markdown auto-invocable

This skill should be used when the user asks about "Zenn記法", "Zenn Markdown", "メッセージブロック", "アコーディオン", "埋め込み", "Zenn特有の記法", or mentions writing Zenn articles with specific formatting needs. Provides comprehensive knowledge of Zenn-specific Markdown syntax and best practices.

Behavior

Syntax reference for Zenn-specific Markdown: message blocks, accordions (:::details), code-block filename/diff extensions, embedded content (link cards, videos, SNS), KaTeX math, mermaid diagrams, and image size/caption controls.

productivity v0.18.0

Productivity helpers for administrative and daily tasks. Install at user scope to make it available across all projects.

Setup

  1. Install at user scope so it is available across all projects.

    /plugin install productivity@atman-marketplace
  2. Install the recommended external skills (grill-me, handoff, writing-great-skills). Requires gh, authenticated.

    /install-recommended-skills
  3. Optional: install the herdr terminal multiplexer and its shell autostart — required by the multiplexer skills (sidekick-go, launch-team, handoff-go, herdr).

    /setup-herdr
  4. The SessionStart plugin-update-check hook needs no configuration.

Typical workflow

Mostly independent day-to-day helpers; the multi-agent skills build on the herdr setup.

  1. Setup

    setup-zellij is the alternative if you prefer Zellij over herdr.

  2. Daily writing and reporting
  3. Multi-agent work inside the multiplexer

    These require running inside herdr (or zellij where supported).

  4. Project documentation
  5. As needed

Skills 17

create-claude-command auto-invocable

Use when building or improving a Claude Code slash command in a project's own `.claude/commands/`: $ARGUMENTS handling, file/bash injection, agent invocation, multi-agent orchestration, or command naming.

Behavior

Reference guide for authoring Claude Code slash commands in a project's .claude/commands/: the scaffold template, $ARGUMENTS/$1 handling, @file and !command injection, naming conventions, and when to use a command vs agent vs skill.

📄 Output

A command markdown file authored in the target project's .claude/commands/ (or ~/.claude/commands/ for personal scope).

create-claude-md auto-invocable

Generate or refine a concise, effective CLAUDE.md for the current project following official best-practice guidelines. Use when the user wants to create, write, or update a project's CLAUDE.md.

Behavior

Inspects the project (build/test/lint config, README, git log conventions) and writes a concise CLAUDE.md that includes only what Claude cannot infer (commands, non-default style rules, workflow conventions, gotchas), routing domain-specific rules to .claude/rules/<domain>.md stubs and keeping the file under ~50 lines where possible.

📄 Output

CLAUDE.md at the project root, plus optional .claude/rules/<domain>.md stubs with paths: frontmatter.

create-readme auto-invocable

Create a comprehensive, well-structured README.md for the current project. Use when the user wants to create, write, or refine a project's README.

Behavior

Investigates the project first (audience/README type, identity, run commands, features, assets, git remote), then writes a README ordered around the dominant intent (consumption, library, or contribution focused) using a prioritized section menu, GFM formatting, verified badges, and GitHub admonitions — never inventing commands or badges it cannot verify.

📄 Output

README.md at the project root.

create-release-notes auto-invocable

Write user-facing release notes from the Conventional Commits history between two refs — grouped by impact, in plain language, with internal-only changes filtered out. Use when the user wants release notes, an announcement of what's new in a version, or a user-readable summary of changes since the last release (distinct from a developer changelog).

Behavior

Collects commits between the last tag and HEAD (or a given range), enriches thin commits from merged PRs via gh, filters out internal-only changes, and writes user-facing notes ordered by impact (Breaking/New/Improved/Fixed) with each item stating the user benefit rather than the implementation.

📄 Output

Draft shown in chat; on request written to a file (convention: docs/releases/<version>.md) or attached to a GitHub release via gh release create/edit.

create-work-log auto-invocable

Compose a daily or weekly work log from git activity across the user's configured repositories and save it as Markdown (e.g. into an Obsidian vault). Use when the user asks for a work log, daily report, weekly summary, 日報/週報, or "what did I do today/this week".

Behavior

Scans the repositories configured in the skill's config.json for git activity in the period (commits on all branches by the configured author, plus uncommitted work), groups it by repo and Conventional Commit theme, asks the user for anything git can't show (or notes auto-generated when unattended), and saves a Markdown log.

📄 Output

<outputDir>/<yyyy-MM-dd>-work-log.md (weekly: <yyyy-MM-dd>-weekly.md), where outputDir comes from the skill's config.json — typically an Obsidian vault folder.

generate-html-report auto-invocable

Generate a rich, self-contained HTML document instead of a long Markdown file — for specs & plans, PR/code-review explainers, design mockups & prototypes, research/status reports, or throwaway editing UIs. Use when the user asks to write a "report", wants a PR or piece of code explained visually, wants to compare design options, needs a research or status writeup, wants a diagram/flowchart, or when a Markdown plan/spec is getting too long to read comfortably.

Behavior

Picks one of five report modes (spec/plan, code-review explainer, design prototype, research/status report, or throwaway editing UI), gathers the real content first (source, git, MCPs, web), and writes one self-contained HTML file with inline CSS/JS/SVG — used in place of any Markdown document that would cross ~100 lines or needs diagrams/interactivity.

📄 Output

A single self-contained .html file saved to a path implied by the mode or chosen by the user, then auto-opened via start/explorer.exe.

handoff-go explicit-only

Write a handoff document and launch the next agent in a split multiplexer pane (herdr or zellij), with new-window and paste-ready fallbacks when no multiplexer is in use.

Behavior

Writes a handoff document summarizing the current conversation (with a suggested-skills section, secrets redacted) to the OS temp directory, then runs the bundled dispatch.mjs which launches a fresh claude session that reads it — splitting a multiplexer pane (herdr/zellij), opening a new terminal window, or printing a paste-ready command as fallbacks. Explicit-invocation only.

📄 Output

Handoff Markdown document in the OS temp directory (not the workspace), consumed by the newly launched session.

Arguments: What the next session should focus on (plus any extra instructions)

herdr auto-invocable requires: Requires the `herdr` CLI on PATH and running inside a herdr-managed pane (HERDR_ENV=1).

Control herdr from inside it. Manage workspaces and tabs, split panes, spawn agents, read output, and wait for state changes — all via CLI commands that talk to the running herdr instance over a local unix socket. Use when running inside herdr (HERDR_ENV=1).

Behavior

Teaches control of the herdr multiplexer from inside it (requires HERDR_ENV=1): listing workspaces/tabs/panes, reading other panes' output, splitting panes and running commands in them, waiting on output or another agent's status, and sending text/keys — all via herdr CLI commands over a local unix socket.

install-skill explicit-only

Install a Claude Code skill from a GitHub skills repository URL.

Behavior

Parses a GitHub skills-repo tree URL (or an OWNER/REPO + path reference) into a repo and skill path, then installs it with gh skill install <repo> <skill-path> --agent claude-code and reports the result. Explicit-invocation only.

launch-team explicit-only requires: Must run inside a supported multiplexer session (herdr or zellij); requires that multiplexer's CLI and the `claude` CLI on PATH.

Launch a role-based team of Claude Code agents in separate multiplexer panes that coordinate on one task by messaging each other directly.

Behavior

Acts as conductor for a multi-agent team: settles 2-4 roles (one orchestrator), writes a self-contained brief and a team config JSON to the OS temp dir, then runs the bundled launcher.mjs which opens one named multiplexer pane per agent (herdr or zellij), writes the shared roster, and delivers the brief to the orchestrator. Requires running inside a multiplexer. Explicit-invocation only.

Arguments: The task plus the roles/number of agents (e.g. "implement feature X with a designer, an implementer, and a reviewer")

manage-desktop-routines explicit-only requires: Requires Claude Code Desktop's scheduled-tasks MCP tools (create_scheduled_task, list_scheduled_tasks, update_scheduled_task).

Create, back up, and restore Claude Code Desktop local routines (scheduled tasks) so they survive a PC change. Use when the user wants to set up a recurring or one-off Desktop routine, back up their routines, or restore routines on a new machine.

Behavior

Three branches for Claude Code Desktop local routines: create (collects taskId/description/self-contained prompt/schedule, calls create_scheduled_task, hands the user a post-creation checklist for folder/model/permissions), backup (dumps all tasks with prompts to a JSON file), and restore (recreates tasks missing on this machine from the backup, never overwriting existing ones). Explicit-invocation only.

📄 Output

<backupDir>/desktop-routines-backup.json, where backupDir comes from the skill's config.json (a synced folder like Drive or an Obsidian vault is recommended).

Arguments: create <what to automate> | backup | restore

post-to-slack explicit-only

Post a message to a Slack channel, resolving channel/user names to IDs and adding an optional @mention or thread reply. Use when the user wants to post, share, or notify something in Slack, or when another skill needs to deliver its output to a Slack channel.

Behavior

Resolves a channel name to its Slack ID via slack_search_channels (and a mention name to <@USERID> via slack_search_users), then posts with slack_send_message (supporting thread replies) and returns the permalink. Intentionally generic so other skills hand it their output instead of calling Slack tools directly. Explicit-invocation only.

setup-herdr explicit-only

Install herdr (terminal-native agent multiplexer), set up its Claude Code and OpenCode integrations, and configure shell profiles to auto-launch herdr on startup (replacing any existing Zellij autostart block). Use when the user wants to install or configure herdr.

Behavior

Checks for herdr in both Windows and WSL (PATH, known install dirs, running process), installs it via the official install scripts only after user confirmation, installs the Claude Code/OpenCode integrations, and edits shell profiles (PowerShell, bash, zsh) to auto-launch herdr — removing any zellij autostart block first. Explicit-invocation only.

setup-zellij explicit-only

Install Zellij and configure shell profiles to auto-launch Zellij on startup (replacing any existing herdr autostart block). Use when the user wants to install or configure Zellij's shell autostart.

Behavior

Checks for zellij in PowerShell and WSL, installs it (winget / apt / cargo) after user confirmation, then edits the four shell profiles to auto-launch zellij on startup, removing any herdr autostart block installed by setup-herdr first. Explicit-invocation only.

sidekick-go auto-invocable requires: Must run inside a supported multiplexer session (herdr or zellij); requires that multiplexer's CLI and the `claude` CLI on PATH.

Use when the user explicitly asks to delegate work to a persistent sidekick/helper agent in its own pane that you can iterate with across multiple rounds (e.g. review -> fix -> re-review) without losing its context each time, or when another skill's instructions explicitly direct invoking sidekick-go. Do not use for a one-off review or one-shot subagent task; that's the Agent/Task tool's job.

Behavior

Spawns a persistent helper agent in an adjacent multiplexer pane (herdr or zellij): writes a self-contained brief to the OS temp dir, runs the bundled dispatch.mjs to open the sidekick's pane and wire two-way messaging over the shared bus, then iterates with it across rounds (replies arrive in the caller's own prompt) until a closing message ends the exchange.

Arguments: The task to delegate, plus (optionally) the sidekick's role, e.g. "Code Reviewer"

wsl-vscode-doctor explicit-only

Diagnose and fix `code .` failing to launch, or opening a disconnected Windows-side VS Code instead of a WSL-connected (Remote-WSL) window, in a herdr-spawned WSL shell.

Behavior

Diagnoses why code . fails or opens a disconnected Windows-side VS Code in a WSL shell (missing PATH entry or missing WSL_DISTRO_NAME), locates the Windows VS Code install, borrows a correct WSL_DISTRO_NAME from a sibling process, and idempotently patches ~/.zshrc with a marker-delimited export block. Explicit-invocation only.

📄 Output

A marker-delimited PATH/WSL_DISTRO_NAME export block appended to ~/.zshrc.

Hooks 1

SessionStart

node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/check-plugin-updates.mjs"

What it does

check-plugin-updates.mjs compares every installed Claude Code plugin (across all marketplaces) against the locally cached marketplace clones and prints the exact update command when a newer version exists — offline, notify-only, and announced once per version per plugin/scope (state in ~/.claude/plugins/.update-notify-state.json).

scrum v0.16.0

Helpers for scrum and agile development workflows, including monday.com backlog and Google Drive doc integration.

Setup

  1. Install the plugin (user scope recommended).

    /plugin install scrum@atman-marketplace
  2. Set your monday.com API token in the shell environment (the bundled CLI scripts and the MCP server both need it).

    setx MONDAY_TOKEN <token>        # Windows (takes effect in new shells)
    export MONDAY_TOKEN=<token>      # WSL/macOS (add to your shell profile)
  3. Install Google Drive for desktop and set the docs folder to mirrored sync — snapshots and reports are written through the local synced folder.

  4. In each project, register the monday board and Epic-to-Drive-folder mappings; writes .claude/scrum-context.json (picked up at the next session start).

    /setup-scrum-context
  5. Optional: pre-install the monday MCP server globally for faster startup; otherwise it falls back to npx.

    npm i -g @mondaydotcomorg/monday-api-mcp

Typical workflow

Skills follow the sprint cycle: one-time setup, per-session context loading, day-to-day backlog work, periodic reporting, and sprint ceremonies.

  1. One-time per project: register the board

    Writes .claude/scrum-context.json (board id, Drive docs root, Epic folder mappings) that every other skill resolves its defaults from.

  2. Start of a session: load the Epic

    Refreshes the backlog snapshot and repo sync when stale, then briefs you from the aggregates at minimal token cost.

  3. Day-to-day backlog work

    Read/update items and keep them sprint-ready (acceptance criteria, estimates, splits).

  4. Periodic snapshots and progress reports

    Typically run from a scheduled routine; both are explicit-invocation wrappers around the bundled scripts.

  5. Sprint ceremonies
  6. As needed

    Spike research, backlog-vs-code audits, spec reports, and Drive doc housekeeping.

Skills 12

audit-epic-consistency auto-invocable

Cross-check an Epic's user stories, PBIs, and acceptance criteria against the actual implementation in its repo, and propose fixes for mismatches, missing/unfinished descriptions, and unclear wording. Use when the user wants a consistency check between backlog and code, 整合性チェック, or asks whether the backlog still reflects what is really implemented.

Behavior

Cross-checks an Epic's backlog against its code: refreshes the snapshot and repo sync if stale, uses .pm/repo/branch-diff.json to scope the investigation to files the Epic actually touched, audits every in-scope item for drift, gaps (missing acceptance.md), and clarity with file:line evidence from the dedicated read-only clone, then applies only user-approved fixes (acceptance.md edits directly, monday edits via monday-api-mcp).

📄 Output

.pm/reports/audits/audit-<YYYY-MM-DD>.md in the Epic's Drive folder.

investigate-spike auto-invocable

Investigate a spike task from an Epic's backlog using the Epic's dedicated repo clone, and save a findings report the team can decide from. Use when the user mentions a spike, スパイク調査, or wants a technical question from the backlog researched against the actual codebase.

Behavior

Identifies the spike item on the monday board, loads its question from the item plus saved updates and acceptance.md, syncs the Epic's dedicated repo clone (sync-repo.mjs), investigates read-only with file:line evidence until there is enough to decide, then writes a findings report and offers to post the conclusion back to the monday item.

📄 Output

Self-contained HTML report at <epicFolder>/.pm/reports/spikes/<itemId>-<slug>.html in the Epic's Drive folder.

load-epic-context auto-invocable

Load a PBL Epic's working context — backlog snapshot, repo sync data, and the Epic summary — refreshing stale data first, so the conversation can start from the Epic's current state. Use when the user wants to load/open an Epic, asks "where are we on <Epic>", Epicをロード, or starts a session about a specific Epic's status.

Behavior

Primes a session with an Epic's state at minimal token cost: refreshes the backlog snapshot (save-all.mjs) and repo sync (sync-repo.mjs) when older than 60 minutes, then reads only the aggregates (summary.md, progress-history.json, repo-state.json, branch-diff.json, tail of commits.jsonl) — never individual item files — and briefs the user before offering to (re)write the summary.

📄 Output

Maintains <epicFolder>/.pm/summary.md (the ~1-page Epic summary) when missing or materially stale.

manage-drive-docs auto-invocable

Search, read, edit, or reorganize (rename/move/delete/list folders) development documentation and its folder structure stored in Google Drive, by accessing the locally synced folder (Google Drive for desktop) directly through the filesystem. Use when the user asks about design docs, specs, or other development documentation kept in Google Drive rather than in the repo, or wants to rename/move/delete files or folders there.

Behavior

Finds, reads, edits, and reorganizes development docs in Google Drive by working directly on the locally synced Google Drive for desktop folder with standard file tools scoped to the configured driveDocsRootPath, falling back to the bundled drive-fs.mjs script for rename/move/delete/list operations and for non-ASCII WSL paths the standard tools cannot reach. Native Google Docs/Sheets formats are out of scope.

manage-monday-backlog auto-invocable

Read or update the product backlog on monday.com. Use when the user asks about backlog items, monday.com boards, tickets, or tasks tracked in monday.com.

Behavior

Reads and updates the monday.com backlog primarily through the plugin's dependency-free CLI scripts (list-items, get-item, update-item-status, save-item/doc/updates/all, init-task, set-link, check-stale-links) which need only MONDAY_TOKEN and Node 18+, resolving the board id and Epic folders from .claude/scrum-context.json; the monday-api-mcp server is the optional path for operations beyond the scripts' surface.

📄 Output

Snapshot scripts write to the Epic's Drive folder under .pm/backlog/ (items/<itemId>.json, docs/<docId>.md, updates/<itemId>.json); init-task.mjs scaffolds PBI markdown files under pbi/.

prepare-sprint-review auto-invocable

Generate a sprint-review HTML document for a PBL Epic — done items with demo pointers, sprint metrics, and an agenda — from the monday.com board and its Drive snapshot. Use when the user wants sprint review material, a demo agenda, or to prepare for a sprint review / スプリントレビュー.

Behavior

Refreshes the Epic snapshot with save-all.mjs, classifies items in the sprint window (done / progressed / carried over), gathers demo material per Done item from its pbi folder's acceptance.md and evidence/ (marking items with no recorded evidence), and generates a self-contained HTML review document with metrics, a demo agenda with time-boxes, per-item cards, and discussion prompts.

📄 Output

<epicFolder>/sprint-reviews/<yyyy-MM-dd>-sprint-review.html in the Epic's Google Drive folder, with its real Drive link resolved via the Drive MCP connector.

refine-backlog auto-invocable

Audit the monday.com backlog for refinement gaps — missing acceptance criteria, missing estimates, oversized or vaguely-titled items — and propose concrete fixes and splits. Use when the user wants backlog refinement/grooming, リファインメント, to check if items are ready for a sprint, or to split a large PBI.

Behavior

Audits every non-Done item (or a narrowed scope) against four readiness checks — acceptance criteria present, estimate present, size within threshold, outcome-stating title — proposes concrete fixes (drafted Given/When/Then criteria, concrete splits, rewritten titles; never invented estimates), and applies only user-approved changes: criteria to the item's acceptance.md, board edits via the monday-api-mcp server or a paste-ready list.

report-epic-spec explicit-only

Generate an HTML report of an Epic's currently-implemented spec and its upcoming planned features, reconciled from the backlog snapshot and the repo's actual state. Use when the user wants a "what exists now vs what's coming" summary, 仕様まとめ, or a feature-status overview for stakeholders.

Behavior

Reconciles what the product does today vs what is planned by reading both sides: board items by status from .pm/backlog/ plus prd/prd.md, and the code via branch-diff.json, recent commits, and targeted reads in the dedicated clone — flagging Done items with no implementation trace and implemented behavior no item records — then writes a stakeholder-readable HTML report. Explicit-invocation only.

📄 Output

<epicFolder>/.pm/reports/spec/spec-report-<YYYY-MM-DD>.html in the Epic's Drive folder.

report-pbl-progress explicit-only

Generate a PBL Epic's progress & completion-forecast HTML report from its Drive snapshot. Use when the user wants a progress/status report, or a burndown/completion forecast for a PBL Epic.

Behavior

Runs the bundled generate-progress-report.mjs once against the Epic's Drive snapshot, which computes totals, status mix, and a completion forecast (assumed velocity until 2+ days of history exist, then measured) and renders the HTML report itself; the skill then resolves the report's Drive link and sharing status via the Drive MCP connector and reports the KPIs — never re-deriving aggregates from item files. Explicit-invocation only.

📄 Output

<epicFolder>/.pm/reports/progress/progress-report-<date>.html, with <epicFolder>/.pm/backlog/progress-history.json upserted one entry per day.

setup-scrum-context auto-invocable

Set up or update .claude/scrum-context.json for a project — the monday.com board, Google Drive docs root, and Epic-to-Drive-folder mappings the scrum plugin's other skills and scripts depend on. Use when the user wants to configure monday/Drive settings for a project, or when manage-monday-backlog/manage-drive-docs report no board id or no Drive root configured.

Behavior

Creates or updates the project's .claude/scrum-context.json without hand-editing: shows current values, extracts the board id from a pasted monday URL, sanity-checks that new mondayEpics folders fall under driveDocsRootPath, then pipes a JSON patch to the bundled init-scrum-context.mjs which merges onto the existing config and ensures the file is git-ignored. Reminds the user a new session is needed for the hook to pick it up.

📄 Output

.claude/scrum-context.json in the project root (merged, and added to .gitignore).

snapshot-pbl-to-drive explicit-only

Snapshot a monday.com PBL Epic's items (PBIs), docs, and updates into its mapped Google Drive folder in one script run.

Behavior

One-shot wrapper that runs save-all.mjs once for the named Epic (board id and Drive folder auto-resolve from scrum-context) and reports the script's JSON summary — deliberately never opening individual snapshot files afterward. Explicit-invocation only.

📄 Output

Bulk snapshot of the Epic's board group into <epicFolder>/.pm/backlog/ (items/, docs/, updates/) plus prd/ in the Epic's Google Drive folder.

write-retrospective auto-invocable

Facilitate and record a sprint retrospective for a PBL Epic — prepare sprint facts from the monday.com board, walk the user through Keep/Problem/Try, save the notes to the Epic's Drive folder, and track action items. Use when the user wants a retrospective, ふりかえり/レトロ, or to record what went well and what to improve after a sprint.

Behavior

Facilitates then records a KPT retrospective: prepares sprint facts from the board (done/carry-over, plus a follow-up verdict on every previous Try action), walks the user through Keep/Problem/Try one pass at a time with data-grounded candidate observations, turns Try items into owned, dated actions, saves the record, and offers to put actions on the monday board via monday-api-mcp.

📄 Output

<epicFolder>/retrospectives/<yyyy-MM-dd>-retro.md in the Epic's Drive folder.

Hooks 1

SessionStart

node "${CLAUDE_PLUGIN_ROOT}/hooks/scripts/inject-scrum-context.mjs"

What it does

inject-scrum-context.mjs reads .claude/scrum-context.json from the project root and injects a <scrum-context> block (monday board id/URL, drive-docs root path(s), and mondayEpics group-to-folder mappings) that the plugin's skills and CLI scripts resolve their defaults from. Silent when the config file is absent.

MCP servers 1

monday-api-mcp

node ${CLAUDE_PLUGIN_ROOT}/mcp/monday-api-mcp-launcher.mjs

Notes

Read/write access to monday.com boards (@mondaydotcomorg/monday-api-mcp) for operations beyond the bundled CLI scripts' surface; requires the MONDAY_TOKEN environment variable (the launcher exits with instructions if unset) and optionally a global npm install for faster startup, otherwise it falls back to npx.

stack-cloudflare v0.1.0

Helpers for developing on Cloudflare (Workers, Pages, R2, D1, etc.).

Setup

  1. Install the plugin; no configuration. Applies when deploying React Router apps to Cloudflare Workers (wrangler).

    /plugin install stack-cloudflare@atman-marketplace

Skills 1

cloudflare-static-assets auto-invocable

Configure and use Cloudflare Static Assets with React Router, ensuring compatibility across development and production environments with automatic fallback handling. Use when deploying React Router applications to Cloudflare Workers that need to access files from the public directory.

Behavior

Guides configuring Cloudflare Static Assets for React Router apps on Workers — adding the assets binding to wrangler.jsonc and creating a utility module that fetches public/ files uniformly across dev (Vite fetch), production (ASSETS binding), and build time (filesystem) — fixing the dev-works/prod-404 mismatch.

stack-dnd-kit v0.1.0

Helpers for building drag-and-drop UIs with dnd kit.

Setup

  1. Install the plugin; no configuration. Applies to React projects using the dnd-kit library.

    /plugin install stack-dnd-kit@atman-marketplace

Skills 1

dnd-kit-implementation auto-invocable

Guide for implementing sortable and droppable components using dnd-kit library. Use this skill when building React applications that require drag-and-drop functionality with both container reordering (useSortable) and item dropping (useDroppable) capabilities, such as Kanban boards, file management systems, or playlist editors.

Behavior

Provides the implementation pattern for combining useSortable and useDroppable on the same container in dnd-kit (Kanban-style UIs): detect what is being dragged via useDndContext and conditionally switch refs/listeners so containers act as drop targets when items drag and as sortables when containers drag.

stack-opencode v0.1.0

Helpers for working with OpenCode configuration and extensions.

Setup

  1. Install the plugin; no configuration. Used when debugging an OpenCode installation.

    /plugin install stack-opencode@atman-marketplace

Skills 1

opencode-log-investigation auto-invocable

Investigates OpenCode runtime errors by locating and analyzing log files. Use when OpenCode fails to start, plugins do not load, MCP servers error out, or terminal shows cryptic errors without details.

Behavior

Diagnoses OpenCode runtime failures by opening the newest log under ~/.local/share/opencode/log/ (or the Windows equivalent) and searching for ERROR lines, with targeted workflows for plugin-loading issues (service=plugin), MCP server failures (service=mcp), session/tool errors, and the gotcha that plugin stdout/stderr never reaches the log file.

stack-react-router v0.1.0

Helpers for React Router and Remix applications.

Setup

  1. Install the plugin; no configuration. Applies when building React Router v7 applications.

    /plugin install stack-react-router@atman-marketplace

Skills 1

react-router-v7-app auto-invocable

Implements React Router v7 app structure, routing patterns, and component templates. Use when creating or modifying React Router v7 applications to ensure consistent folder structure, data loading patterns, and component architecture.

Behavior

Prescribes React Router v7 conventions: file-based routing in app/routes/ (folder routes, route groups, dot-notation nesting, excluding test files from route discovery), loader/action data patterns with typed Route.* props, arrow-function route components, ErrorBoundary usage, and meta/headers exports.