Skip to content

v2.32.0

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 14 Aug 23:43
· 56 commits to main since this release
v2.32.0
e8a641c

Project memory: the catalog stops asking you to repeat yourself.

Every previous release added capability you invoke. This one adds capability that accumulates.

A skill invocation has always started from zero: you paste the same personas, restate the same initiative, re-explain which phase you are in. v2.32.0 introduces an opt-in state file that eight skills read, so the second skill in a session knows what the first one produced.

The concrete moment: synthesize your research with discover-interview-synthesis, then run deliver-prd and it uses the personas you already produced instead of asking you to paste them again.

Nothing happens until you opt in. With no .claude/pm-skills.local.md present, every skill and both hooks behave exactly as they did in v2.31.1. Writes are proposed for your confirmation rather than applied, and the flag that makes them automatic fails closed.

No new skills. Catalog stays 68 skills (30 phase + 11 foundation + 12 utility + 15 tool) and 6 sub-agents. Additive MINOR.

Also in this release

  • Trigger-eval coverage closes as an accounting, not a percentage: 53 skills carry a fixture pack and 15 are excluded by design, covering all 68 with nothing unclassified, now asserted as real set equality against the catalog rather than a hardcoded count.
  • Release automation gains the three mechanisms its shadow-to-authoritative cutover was missing. This cycle produced the first fully green Release PR in the program's history.
  • A tracked Phase-0 scope spec for a future AI-product skill family, with a new build gate: a proposed skill must beat a no-skill control arm before it is built. Nothing in that family is built here.

Two defects worth naming

An adversarial review before the tag found two claims verified against the thing that produced them rather than the thing a user meets.

The published project-memory config example did not parse: it carries inline YAML comments the frontmatter reader did not strip, so anyone copying it verbatim got no declared phase and no warning. And the coverage claim above was not actually asserted, only asserted-about in a comment.

Both are fixed here. They are called out because the failure mode generalizes: a test written by the same hand that wrote the thing it tests can pass while the claim it certifies is false.

Full release notes: https://product-on-purpose.github.io/pm-skills/releases/Release_v2.32.0/
Changelog: https://github.com/product-on-purpose/pm-skills/blob/main/CHANGELOG.md

What's Included

  • PM skills library (flat skills/{phase-skill}/ + foundation capabilities)
  • Sample output library (library/skill-output-samples/ with README and canonical sample corpus)
  • Slash commands (10 /workflow-* orchestrators plus the /chain ad-hoc runner; every skill invocable by name)
  • Workflows (Triple Diamond, Lean Startup, Feature Kickoff)
  • Claude plugin manifest (.claude-plugin/plugin.json)
  • Claude/openskills discovery via sync helper (scripts/sync-claude.sh / .ps1)
  • Docs: README, QUICKSTART, AGENTS, reference guides

Download

pm-skills-v2.32.0.zip - Complete package for all platforms

Installation

Claude Code plugin (recommended)

/plugin marketplace update
/plugin install pm-skills@product-on-purpose

Claude Code / openskills (zip)

  1. Download pm-skills-v2.32.0.zip
  2. Extract, then run:
    • Bash: ./scripts/sync-claude.sh
    • PowerShell: ./scripts/sync-claude.ps1

Verify the download

sha256sum -c pm-skills-v2.32.0.zip.sha256
Assets 5
Loading