Proposal: New pm-governance plugin — decision governance layer (4 skills, 3 commands)

[AnshulKummar]

Summary

Proposing a new pm-governance plugin that adds a decision governance layer on top of the existing pm-skills ecosystem. While your plugins handle PM execution (what to do), this plugin handles when to pause and how to decide — preventing the most expensive PM mistake: building the wrong thing at full speed.

Proposed Plugin: pm-governance

4 Skills

Skill	Description	Auto-invokes when...
clarification-gate	Asks max 5 clarifying questions before any build; states assumptions first, then confirms	User says "build", "implement", "create", "design"
honest-evaluator	6-part evaluation framework: strengths (specific mechanisms only), fatal flaws, hidden costs, risk heatmap, verdict (BUILD NOW / LATER / DIFFERENTLY), alternative	User says "should we build", "evaluate", "pros and cons"
adr-creator	Generates Architecture Decision Records with cost analysis at 4 scale tiers (MVP / 10K / 1M / 100M users)	User mentions database, infrastructure, scaling, "X or Y"
instruction-auditor	Detects anti-patterns in CLAUDE.md and skills (>200 lines, linting rules in CLAUDE.md, vague descriptions) and suggests specific fixes	User asks to review setup or output quality is declining

3 Commands

Command	Description	What it chains
/act [goal]	Structured 5-Act experience: Context → Discovery → Design (pause) → Implementation → Review	clarification-gate → /evaluate → ADR → build → /update-backlog
/evaluate [idea]	Brutally honest evaluation — no hedging, no flattery	honest-evaluator skill
/update-backlog	Syncs TASKS.md: marks completed, adds discovered items, logs parked items with reasons	Session review

How it complements existing plugins

Existing pm-skills                    pm-governance adds
─────────────────────                 ─────────────────────
/discover → what to explore           clarification-gate → what to ask first
/write-prd → what to document         /evaluate → whether to build at all
/strategy → what direction            adr-creator → which technology + cost
/sprint → what to build               /act → structured flow with pause gate
/plan-launch → how to ship            /update-backlog → what to track after

The /act command is designed to chain naturally with existing pm-skills commands — it suggests /discover in Act 2 and /write-prd in Act 3.

Proposed directory structure

Follows the existing plugin pattern exactly:

pm-governance/
├── .claude-plugin
├── README.md
├── commands/
│   ├── act.md
│   ├── evaluate.md
│   └── update-backlog.md
└── skills/
    ├── adr-creator/SKILL.md
    ├── clarification-gate/SKILL.md
    ├── honest-evaluator/SKILL.md
    └── instruction-auditor/SKILL.md

No changes to any existing files. Zero merge conflicts.

Additional context

I built these as part of a companion playbook for PMs using Claude Code: ai-pm-playbook. The playbook also includes hooks (skill-router, safety guards, backlog automation) and a full 16-chapter guide — but only the skills and commands above are relevant for a plugin contribution.

Happy to submit a PR with the full plugin if you're interested, or adjust the scope based on your feedback. The skills and commands follow your existing frontmatter format (name, description for skills; description, argument-hint for commands).

About me

I'm Anshul Kumar — Principal Group Product Manager at Paylocity, focused on platform/API products, integrations, and AI. IEEE Senior Member with published research on LLM frameworks (IEEE AIxSET). Advisory Board Member at Products That Count.

• LinkedIn: https://www.linkedin.com/in/anshula42/
• GitHub: https://github.com/AnshulKummar
• Playbook repo: https://github.com/AnshulKummar/ai-pm-playbook

Thank you for building pm-skills — it's the foundation this work builds on.

[dempseydata]

Good suggestion