Documentation orchestrator for AI-assisted development. Generates, audits, and maintains project documentation through parallel agent dispatch, agentic workflow discovery, Mermaid diagram generation, and formal spec lifecycle tracking.
A superset of obra/superpowers documentation patterns — extends them with automated discovery of agentic pipelines (skills, commands, MCP tools), multi-scope parallel auditing, architecture diagram generation, formal specification tracking through implementation, workflow hooks for automated freshness monitoring, and release notes management.
doc-superpowers is a Claude Code skill that treats documentation as a first-class engineering artifact. It:
- Discovers your project's doc tooling, directory structure, and agentic workflows automatically
- Generates a complete documentation suite from scratch (
init) - Audits existing docs against current code for staleness (
audit) - Reviews PR-scoped documentation impact (
review-pr) - Updates stale docs with agent-verified changes (
update) - Regenerates architecture and workflow diagrams (
diagram) - Syncs doc indexes with the filesystem (
sync) - Installs opt-in workflow hooks for automated freshness monitoring (
hooks) - Tracks specifications through implementation with formal spec lifecycle (
spec-generate,spec-inject,spec-verify) - Drafts release notes from git history with agent-assisted diff review (
release) - Syncs CLAUDE.md and README.md automatically across all write actions to prevent drift
- Tracks freshness via bundled
scripts/doc-tools.sh— content hashing for docs, commit SHA comparison for code
Copy or symlink into your Claude Code skills directory:
# Clone
git clone git@github.com:woodrowpearson/doc-superpowers.git ~/code/doc-superpowers
# Symlink into Claude Code skills
ln -s ~/code/doc-superpowers ~/.claude/skills/doc-superpowersCopy SKILL.md and references/ into .claude/skills/doc-superpowers/ in any project.
Use Cursor's plugin system:
/add-plugin doc-superpowers
Or clone and point .cursor-plugin/plugin.json at the repo.
git clone https://github.com/woodrowpearson/doc-superpowers.git ~/.codex/doc-superpowers
mkdir -p ~/.agents/skills
ln -s ~/.codex/doc-superpowers ~/.agents/skills/doc-superpowersSee .codex/INSTALL.md for details.
Add to your opencode.json:
{
"plugin": ["doc-superpowers@git+https://github.com/woodrowpearson/doc-superpowers.git"]
}See .opencode/INSTALL.md for details.
gemini extensions install https://github.com/woodrowpearson/doc-superpowersnpx skills add woodrowpearson/doc-superpowersWorks with 40+ supported agents. See skills.sh for details.
/doc-superpowers <action> [scope]
Actions: init | audit | review-pr | update | diagram | sync | hooks | release | spec-generate | spec-inject | spec-verify
Scopes: all | <auto-detected from docs/ structure>
| Action | Purpose | When to Use |
|---|---|---|
init |
Generate full doc suite from scratch | New project or missing docs |
audit |
Check all docs for staleness | Periodic health check |
review-pr |
Check docs affected by PR changes | Before merging PRs |
update |
Apply fixes from audit/review | After audit identifies stale docs |
diagram |
Regenerate architecture diagrams | After structural changes |
sync |
Sync doc index with filesystem | After adding/removing doc files |
hooks |
Install workflow hooks (git, Claude Code, CI/CD) | Setting up automated freshness monitoring |
release |
Draft release notes entry from git history | Cutting a new version |
spec-generate |
Generate formal specs from design doc | After brainstorming produces a design spec |
spec-inject |
Inject spec tasks into plans, track during execution | During plan writing and after each chunk executes |
spec-verify |
Verify spec compliance, review spec coverage | Before merging or during code review |
# Generate docs for a new project
/doc-superpowers init
# Audit all documentation (writes report to docs/plans/)
/doc-superpowers audit
# Check docs before merging a PR
/doc-superpowers review-pr
# Regenerate diagrams
/doc-superpowers diagram
# Draft release notes from git history
/doc-superpowers release
# Override the starting commit
/doc-superpowers release --from=v2.2.0# Generate formal specs from a design doc
/doc-superpowers spec-generate --design-doc=docs/superpowers/specs/2026-03-14-feature-design.md
# Inject spec tasks into an implementation plan
/doc-superpowers spec-inject --phase=plan --plan=docs/superpowers/plans/2026-03-14-feature.md --specs=docs/specs/SPEC-AUTH-001-oauth-flow.md
# Check spec freshness after a chunk executes
/doc-superpowers spec-inject --phase=execute --specs=docs/specs/SPEC-AUTH-001-oauth-flow.md
# Final compliance check before merging
/doc-superpowers spec-verify --mode=post-execute --specs=docs/specs/SPEC-AUTH-001-oauth-flow.md --design-doc=docs/superpowers/specs/2026-03-14-feature-design.md
# Spec coverage check during review
/doc-superpowers spec-verify --mode=review --changed-files=src/auth/oauth.py,src/auth/session.pyFor wrapper skill integration, see references/spec-lifecycle-protocol.md.
Install opt-in hooks for automated freshness monitoring:
# Install all hook tiers
/doc-superpowers hooks install --all
# Or pick specific tiers
/doc-superpowers hooks install --git # Git hooks
/doc-superpowers hooks install --claude # Claude Code hooks
/doc-superpowers hooks install --ci # GitHub Actions
# Check what's installed
/doc-superpowers hooks status
# Remove hooks
/doc-superpowers hooks uninstall --allGit hooks: Pre-commit warns when staged files affect stale docs. Post-merge and post-checkout alert on branch switches. Prepare-commit-msg injects freshness comments.
Claude Code hooks: Pre-commit gate catches Claude-initiated commits. Session summary reminds about stale docs when ending a session.
CI/CD: PR freshness check comments on PRs. Weekly cron detects drift. Post-merge workflow keeps the doc index in sync.
Set DOC_SUPERPOWERS_STRICT=1 to make pre-commit block instead of warn. Set DOC_SUPERPOWERS_QUIET=1 to suppress hook output while still enforcing checks. Set DOC_SUPERPOWERS_SKIP=1 to bypass all hooks temporarily.
The init action generates a structured documentation suite in docs/:
| Directory/File | Content | When Generated |
|---|---|---|
architecture/system-overview.md |
System overview, C4 diagrams, tech stack | Always |
architecture/{component}.md |
Per major component/domain | application scope |
architecture/diagrams/ |
C4, component, ERD diagrams | Always |
specs/README.md + template.md |
Spec index and template | Always |
adr/README.md + template.md |
ADR log and template | Always |
workflows/{name}.md |
Process flows, CI/CD | Always |
workflows/agentic/{skill}.md |
Agentic workflow docs | agentic scope |
workflows/diagrams/ |
Workflow, sequence, state diagrams | Always |
guides/getting-started.md |
Prerequisites, installation, verification | Always |
api-contracts.md |
Endpoints, schemas, request/response | api-contracts scope |
data-layer.md |
Data models, ERD, storage | data-layer scope |
ci-cd.md |
Pipeline overview, triggers, environments | ci-cd scope |
infra.md |
Infrastructure topology, components | infrastructure scope |
codebase-guide.md |
Directory map, key files, code flow | Always |
conventions.md |
Code style, naming, git conventions | Always |
.doc-index.json |
Machine-readable freshness index | Always |
doc-superpowers automatically discovers Claude Code artifacts that define agentic pipelines:
- Skills (
.claude/skills/*/SKILL.md) — sub-agents, scripts, user gates - Commands (
.claude/commands/*.md) — which skills they invoke - MCP tools (MCP config files) — server names and tool purposes
- Scripts (
scripts/) — roles in pipelines (dispatch, validate, merge)
Each discovered workflow gets documented with:
- Pipeline overview flowchart
- Phase/session subgraph diagrams
- Multi-actor sequence diagrams with sub-agent lifelines
- State diagrams for pipelines with recovery flows
| Level | Meaning |
|---|---|
| P0 Critical | Doc describes behavior code no longer implements |
| P1 Stale | Code changed, doc probably needs updating |
| P2 Incomplete | Doc missing sections for new functionality |
| P3 Style | Formatting, broken links, outdated terminology |
doc-superpowers uses a hub-and-spoke architecture:
- Discovery phase runs first, building an inventory of the project
- Action router dispatches to the requested action
- Parallel agents handle scope-isolated reviews (one agent per doc scope)
- Verification gates ensure agent findings include evidence (exact doc vs code quotes)
- Output is a structured report with severity-ranked findings
This skill is designed as a documentation superset of the obra/superpowers framework:
- Uses the same skill structure conventions (SKILL.md frontmatter, description triggers)
- Follows superpowers' verification-before-completion patterns
- Extends with documentation-specific workflows not covered by the base framework
- Compatible with superpowers' code review integration (callback pattern)
doc-superpowers/
├── .gitignore # Git ignore rules
├── .claude-plugin/ # Claude Code plugin manifest + marketplace
│ ├── plugin.json
│ └── marketplace.json
├── .cursor-plugin/ # Cursor plugin manifest
│ └── plugin.json
├── .codex/ # Codex installation guide
│ └── INSTALL.md
├── .opencode/ # OpenCode plugin + installation guide
│ ├── INSTALL.md
│ └── plugins/
│ └── doc-superpowers.js
├── SKILL.md # Main skill definition
├── AGENTS.md # Cross-client agent instructions
├── GEMINI.md # Gemini CLI context redirect
├── gemini-extension.json # Gemini CLI extension manifest
├── package.json # npm/OpenCode package metadata
├── scripts/
│ ├── doc-tools.sh # Bundled freshness tooling
│ ├── test-doc-tools.sh # Test suite for doc-tools.sh
│ ├── test-helpers.sh # Shared test utilities
│ ├── test-hooks.sh # Test suite for hooks installer
│ └── hooks/
│ ├── install.sh # Hook installer engine
│ ├── git/ # Git hook scripts (pre-commit, post-merge, etc.)
│ ├── claude/ # Claude Code hook scripts
│ └── ci/ # GitHub Actions workflow templates
├── references/
│ ├── doc-spec.md # Templates and conventions
│ ├── agent-prompt-template.md # Review agent prompt template + scope focus areas
│ ├── output-templates.md # Audit report format + plan template
│ ├── spec-lifecycle-actions.md # Detailed procedures for spec lifecycle actions
│ ├── spec-lifecycle-protocol.md # Spec lifecycle integration guide
│ ├── integration-patterns.md # Code review, commit review, wrapper skill integration
│ └── tool-mappings.md # Cross-framework tool name mappings
├── evals/ # Evaluation test cases
│ └── evals.json # Test prompts and assertions
├── docs/ # Documentation about this skill
│ ├── architecture/
│ │ ├── system-overview.md
│ │ └── diagrams/
│ ├── workflows/
│ │ ├── doc-superpowers.md
│ │ └── diagrams/
│ ├── guides/
│ │ └── getting-started.md
│ ├── superpowers/
│ │ ├── specs/ # Design specs from brainstorming
│ │ └── plans/ # Implementation plans from writing-plans
│ ├── .doc-index.json # Machine-readable freshness index
│ ├── plans/ # Audit reports and update plans
│ ├── archive/ # Archived docs
│ ├── codebase-guide.md
│ └── conventions.md
├── README.md
├── LICENSE # MIT
├── RELEASE-NOTES.md
└── CLAUDE.md
The skill itself (SKILL.md + references/) has zero dependencies. The bundled tooling in scripts/ requires:
| Dependency | Required | Notes |
|---|---|---|
git |
Yes | Already required by doc-superpowers |
jq |
Yes | brew install jq / apt install jq |
sha256sum or shasum |
Yes | Standard on Linux/macOS respectively |
- Fork the repository
- Create a feature branch
- Make changes to
SKILL.mdorreferences/doc-spec.md - Test with
/doc-superpowers initon a sample project - Submit a PR
MIT License. See LICENSE.