Mission: Turn any codebase into professional, marketing-ready repository documentation — powered by AI coding assistants.
PitchDocs is a pure Markdown plugin with 15 skills, 4 agents (3 pipeline + 1 per-project freshness checker), 1 auto-loaded rule + 2 installable rules, 16 commands (14 active + 2 stubs), and 21 evaluation test cases. This roadmap outlines completed work, current priorities, and future directions.
Target: Q1 2026
- Add
/pitchdocs:api-referencecommand (#41) — Direct slash command for TypeDoc, Sphinx, godoc, and rustdoc integration. Currently unreachable through primary/pitchdocs:*pattern. Needs command definition, 1–2 eval test cases, and skill cross-references touser-guidesandplatform-profiles - Resolve upstream spec version drift (#12) — Contributor Covenant v3.0 detected in monthly check (latest is v2.1). Update
upstream-versions.jsonand refresh skill content
- ✅ Auto-loaded rules reduced to ~500 tokens (moved advisory features to per-project install)
- ✅ 4 of 6 over-budget skills split into companion reference files
- ✅ Per-project activation system (
/pitchdocs:activate) for doc-standards, docs-awareness, docs-freshness agent, and optional content-filter hook - ✅ Auto-loaded context reduced by 41%,
/readmecontext overhead reduced by 72% for small projects
- Completed per-project activation system (
/pitchdocs:activate install/install strict) - Installable rules:
doc-standards.md,docs-awareness.mdmoved from.claude/to source templates - Installable agent:
docs-freshness.mdread-only freshness checker (per-project) - Strict tier: optional
content-filter-guard.shhook for Claude Code - Fully backwards-compatible — all 15 commands work globally regardless of activation tier
- Breaking: Split AI context management into ContextDocs — stub redirects remain for
/pitchdocs:ai-contextand/pitchdocs:context-guard - Added 6 automated CI checks (spell check, frontmatter validation, llms.txt consistency, banned phrases, orphan detection, token budgets)
- Added skill activation eval framework with 21 test cases — 95.2% accuracy on Haiku
- Split 4 over-budget skills into companion reference files for token budget compliance
- Added comprehensive documentation: getting-started tutorial, workflow guides, launch artifacts, ROADMAP
- Integrated website notification into release-please workflow
- Resolved Context Guard false positives and added Untether session detection
- Replaced demo screenshot TODO with showcase gallery
- Added website accuracy audit against v1.19.1
- Reduced
/readmecontext overhead by 72% for small projects - Regenerated llms-full.txt with slimmed skills and reference files
- Reduced auto-loaded context by 41% and fixed stale counts across docs
- Resolved false-positive broken path warnings in drift check hook
- Added dark mode logo support and demo screenshot placeholder
- Extracted on-demand skills from auto-loaded rules for 56% context token reduction (#28)
- Added Signal Gate principle, lifecycle commands, and context health scoring (#25)
- Added auto-memory (MEMORY.md) accommodations to AI context guidance (#24)
- Added root-level SKILL.md for directory submissions
- Added two-tier context doc enforcement to Context Guard
- Added user benefits extraction with persona inference and docs-awareness rule
- Added GitLab and Bitbucket platform support
- Auto-update docs on release — trigger
/pitchdocs:doc-refreshwhen version tags are created - GitHub Actions template for scheduled doc refreshes
- Enhanced content filter mitigation — reduce need for chunked writes on legal templates
- Skill quality benchmarking in CI — track A/B comparison scores over time
- Multi-language README support — generate localised docs for i18n projects
- Interactive README builder — REPL-style step-by-step Q&A for custom documentation
- Blog post generator — auto-extract Dev.to, Medium, Hashnode-ready content from README
- Jira / Linear integration — pull roadmap data from issue tracking platforms
- API reference direct integration — complete
/pitchdocs:api-referencewith API doc linking
- Gitea, Forgejo, and Sourcehut platform support
- GitHub discussion template generator
- Expand user benefits extraction to industry-specific personas (DevOps, ML ops, data science)
- Integration tests for all 9 supported AI tools (Claude Code, OpenCode, Codex, Cursor, Gemini, Windsurf, Cline, Aider, Goose)
Found a bug or have a feature request? Open an issue — all feedback is valuable.
Looking for your first contribution? Check out good first issues — these are scoped for new contributors and have clear acceptance criteria.
- Help test PitchDocs on your own projects — report edge cases or improvements
- Contribute to guides for other AI tools
- Improve examples in skills or commands
Interested in deeper involvement? See CONTRIBUTING.md for the full contributor guide, code review process, and commit conventions.
| Metric | Current | Target |
|---|---|---|
| Plugin version | 2.1.0 | 3.0.0 (2026) |
| Auto-loaded context | ~500 tokens | <1,000 tokens |
| Skills exceeding 2k token limit | 2 | 0 |
| Supported platforms | 3 (GitHub, GitLab, Bitbucket) | 6+ by v4.0 |
| AI tool compatibility | 9 (including stubs) | 12+ |
| Activation eval test coverage | 21 scenarios (95.2% Haiku) | 30+ |
| Generated doc types | 20+ | 25+ |
- Getting started: See Getting Started Guide
- Troubleshooting: See Troubleshooting Guide
- Contribution: See CONTRIBUTING.md
- Support: See SUPPORT.md
Last updated: 2026-03-14