GitHub Planner is a PM toolkit built as an MCS techpack. If you've managed backlogs, triaged issues, or tracked delivery and found something missing — your contribution is welcome.
Add a new type to /plan-discussion. The analysis type is a good reference — it's intentionally agnostic. A template needs:
- Interview questions (in
src/skills/discussion-builder/SKILL.md) - Markdown template (in
src/skills/discussion-builder/references/) - Category mapping for GitHub Discussions
- Entry in the Knowledge Manifest schema
Add a new --scope to /plan-insights. The ci scope is a recent example. A scope needs changes in 4 files:
- Skill (
src/skills/plan-insights/SKILL.md) — metrics, thresholds, decision tips - Agent (
src/agents/plan-insights-analyst.md) — fetch commands and computation steps - Report template (
src/skills/plan-insights/references/plan-insights-report.md) — output format - Command (
src/commands/plan-insights.md) — scope option and report section
Better heuristics for:
- Label mapping (document signals → GitHub labels)
- Discussion type detection (content signals → type slug)
- Triage rules (priority re-evaluation, staleness, duplicates)
- Pipeline stage detection (PR-to-issue linking)
Fixes, typos, clearer examples, or better --help-style descriptions in commands.
We use a fork-based workflow. All contributions come through pull requests from your fork.
gh repo fork anettodev/github-planner --clone
cd github-plannergit checkout -b <type>/<short-description>Branch naming:
feature/— new capability (template, scope, detection rule)fix/— bug fixdocs/— documentation only
See Techpack Structure below to know where things go.
# In any project where you want to test:
mcs pack add /path/to/your/github-planner
mcs sync
mcs doctorVerify:
mcs doctorpasses with no issues- Your new command/scope runs without errors
- Existing commands still work (no regressions)
git push origin <your-branch>
gh pr createInclude in your PR description:
- What — what you added or changed
- Why — the use case or problem it solves
- How to test — steps to verify it works
src/
skills/ → Knowledge and rules (what to do)
agents/ → Execution logic (how to do it)
commands/ → User entry points (slash commands)
config/ → Settings and permissions
hooks/ → Session lifecycle scripts
templates/ → CLAUDE.local.md content
techpack.yaml → Pack manifest
| If you're adding... | You'll touch... |
|---|---|
| A discussion type | src/skills/discussion-builder/ + references/ + src/agents/discussion-manager.md + src/commands/plan-discussion.md |
| An insight scope | src/skills/plan-insights/ + references/ + src/agents/plan-insights-analyst.md + src/commands/plan-insights.md |
| A triage rule | src/skills/issue-triage/ + src/agents/issue-analyst.md |
| A label mapping | src/skills/plan-to-issues/SKILL.md |
A new gh command |
src/config/settings.json (add to permission allowlist) |
Always update README.md if your change adds user-facing functionality.
- Write for PMs, not developers. Plain language over jargon.
- Every report needs color indicators. Use the severity scale: 🟢 healthy, 🟡 warning, 🟠 stale, 🔴 critical.
- Include decision tips. Don't just show data — tell the PM what to do about it.
- Read-only by default. Any mutation (create, close, label) requires explicit flags like
--apply. - Sparklines over tables when possible.
▂▄▅▇▅▃conveys trends faster than a column of numbers. - Keep templates agnostic. Don't hard-code platform-specific assumptions (e.g., iOS-only, App Store-only).
Be kind. Project management is stressful enough. We're all trying to make backlogs less painful.