Releases
All PM Skills releases with detailed notes on what changed and why.
| Version | Date | Highlights |
|---|---|---|
| v2.31.0 | 2026-07-05 | The zero-drift release, generation becomes the only write path for the README catalog, quickstarts, compat matrix, and manifest descriptions; release-please goes live in shadow; the two awk-hazard validators are ported to single-source Node; trigger-fixture coverage climbs to 43 of 68 skills; and SECURITY.md gains a provenance page. |
| v2.30.0 | 2026-07-04 | A trust-repair release that fixes count drift on the front-door install surfaces, closes the CI gate holes that let it pass, stages agents/ and hooks/ in the release zip, and sweeps headings, descriptions, and tool-family versioning. |
| v2.29.1 | 2026-06-24 | A maintenance patch that fixes silently-dropped sections on skill documentation pages - the site generator now renders every SKILL.md section, guarded by an output-based CI gate. |
| v2.29.0 | 2026-06-23 | A new foundation skill, build-risk-review, that names the one assumption most likely to make a build fail and returns a verdict before you commit, plus pm-skill-router, a key-free router that makes the collision gate runnable on the subscription. |
| v2.28.0 | 2026-06-20 | New foundation skill foundation-stakeholder-briefings: fans any source artifact (spec, discovery, research, GTM, experiment, retro) into one canonical master document plus a set of audience-tailored briefings, one per stakeholder lens (executive, board, engineering, UX, PMM, sales, CS, legal, data, or a custom audience). Master-first projection: each briefing cites the master claim IDs it draws on and carries exactly one ask, so the versions never disagree; an advisory check-briefings-trace validator enforces it. Nine first-class lenses with boundaries + an overlap matrix; source-aware audience proposal; 18 library samples. Catalog grows 66 to 67 (foundation 9 to 10); additive MINOR. |
| v2.27.1 | 2026-06-16 | Maintenance patch: closes the classification sub-count drift class. check-count-consistency now derives the four per-classification / per-phase skill sub-counts (phase, foundation, utility, tool) from frontmatter and validates them on every doc surface, so a stale “10 utility skills” or “Foundation Skills (9)” fails CI like a stale total - the gap that let the classification split drift on a published page until a v2.27.0 hand audit caught it. Additive (the total-count check is unchanged); historical mentions use word-form or count-exempt. Three gate-surfaced stale sub-counts fixed. No skill behavior change; catalog stays 66. PATCH. |
| v2.27.0 | 2026-06-15 | The provable-quality release. Makes skill quality verifiable and regression-protected with no new skills (catalog stays 66). Trigger-accuracy evals: every measured skill carries labeled evals/trigger-fixtures.json; a controlled router eval scores per-skill recall/precision against a committed baseline; CI gates fail on routing drift or a new-skill collision. Generated catalog surfaces: skill-manifest.json and the AGENTS.md catalog are built from frontmatter behind enforcing staleness gates, retiring the hand-sync drift class. Output-quality eval infrastructure (harness, six family rubrics, the tested aggregation + verdict module, a three-arm informed control, the asset gate) ships with per-skill results kept internal. The creator/validator family bakes the eval contract into skill creation, and the reciprocity gate is enforcing. Additive MINOR. |
| v2.26.0 | 2026-06-10 | The authoring and quality release. The /chain command runs any ad-hoc ordered skill sequence through the existing pm-workflow-orchestrator engine (Mode B, checkpointed by default; grammar written once as the Mode B Chain Expression Contract; no new engine, no new skill), and the new utility-pm-workflow-builder skill turns a proven chain or a fresh idea into a staged draft workflow packet (workflow file + command + promotion checklist) for human review. The quality-convergence effort completed: all 26 original-generation skills gained “When NOT to Use” boundary pointers and enumerated output contracts, zero instruction rewrites, zero template changes. The orchestrator’s native delegation path was live smoke-tested on the installed plugin before tagging (recorded PASS; procedure now a repeatable runbook). Catalog grows 65 to 66 (utility 11 to 12); command files 10 to 11; additive minor. |
| v2.25.2 | 2026-06-10 | Maintenance patch: resolves the remaining 2026-06-06 Codex audit items and hardens the release gate. A single validator-inventory manifest (scripts/validation-manifest.yaml) plus an enforcing CI parity referee (scripts/check-validator-parity.mjs) so the bash, PowerShell, and CI validator inventories cannot drift (it caught and reconciled a live drift); the root-document link checker extended to source files; the em-dash-scar guard promoted to enforcing and multi-backtick aware; and the CLAUDE.md gitignored-claim correction. A Codex adversarial review ran first; findings resolved before tagging. No skill behavior change; catalog stays 65 skills / 5 sub-agents; patch. |
| v2.25.1 | 2026-06-06 | Maintenance patch: banks accumulated untagged work since v2.25.0. Documentation-site reorg to the Product on Purpose family layout (Pattern S) with full family-standard conformance; a generated, CI-gated resource index (docs/RESOURCES.md) plus a docs/ front door; root-document link repair after the relocation with a new enforcing CI guard; an em-dash-scar cleanup (residual . swept to -) across user-facing and internal prose with new advisory and enforcing guards; three site dependency bumps; and a pre-tag validator parity fix reconciling the PowerShell bundle with bash and CI. No skill behavior change and no published-URL change (route parity verified). Catalog stays 65 skills / 5 sub-agents; patch. |
| v2.25.0 | 2026-06-03 | Activation and trust layer: the plugin’s first hooks. Opt-in house-rule guardrails (a PreToolUse hook that blocks em-dash and en-dash characters when guardrails: true is set in .claude/pm-skills.local.md; placeholder and metric checks warn) and a confident-only phase router (a SessionStart hook that suggests the right Triple Diamond skills only when a repo signal is strong, and stays silent otherwise). Adds M-30, an advisory output-quality CI tier: three deterministic invariant validators over the recorded samples (no placeholders, exact-quote sourcing, no fabricated metrics). Hooks are dependency-free and fail open; hooks are a Claude Code primitive. No new skills (catalog stays 65; sub-agents stay 5); additive minor. |
| v2.24.0 | 2026-06-01 | Workflow orchestrator (the promised v2.17 component, shipped): new utility-pm-workflow-orchestrator dispatch skill + pm-workflow-orchestrator sub-agent run a saved foundation-prioritized-action-plan (or a named chain) through its recommended pm-skills, checkpointing for go/no-go by default and refusing to advance past a failed or empty step. foundation-prioritized-action-plan gains a --run handoff to v1.1.0 (one confirmation hands the plan to the governed orchestrator; the plan skill still never runs work-skills inline). Ships EXPERIMENTAL on every client, native Claude Code path included, until live smoke tests close; --dry-run is the safe preview. Catalog grows 64 to 65 (utility 10 to 11); fifth sub-agent pm-workflow-orchestrator; foundation-prioritized-action-plan to v1.1.0; additive minor. |
| v2.23.0 | 2026-05-31 | New foundation skill foundation-prioritized-action-plan: turns any PM input (notes, transcripts, drafts, executive asks, raw situations) into one evidence-grounded prioritized action plan. Theory of Constraints ranks by the single binding constraint; Cynefin caps confidence (probes for Complex, stabilization for Chaotic). A source ledger is built before analysis and every claim cites an exact input quote, so it refuses High-confidence plans for uncertain situations. Catalog grows 63 to 64; additive minor, no existing skill changed. |
| v2.22.0 | 2026-05-30 | Wrapper deletion + native Codex support: the 63 duplicate per-skill command wrappers are removed so each capability appears once in the menu (commands 73 to 10, keeping the 10 /workflow-* orchestrators); a native .codex-plugin/plugin.json manifest is added so Codex discovers the skills (it previously reported “No plugin skills”). All 63 skills unchanged; shipped as a minor. |
| v2.21.0 | 2026-05-26 | Marketplace launch (additive): pm-skills is now published through the new product-on-purpose marketplace, a single home for multiple Product on Purpose plugins. Distribution change only - your skills, commands, and their behavior are identical (catalog 63). The previous install path keeps working, so existing installs need no action; shipped as a minor. |
| v2.20.0 | 2026-05-25 | Sprint Workflow Commands + validation hardening: the three workshop methodologies become single slash commands (/workflow-foundation-sprint, /workflow-design-sprint, /workflow-foundation-to-design) chaining their per-day skills end to end; slash commands 70 to 73, no new skills (catalog 63). Command-sync and count-consistency validators tightened; the near-vestigial bundle-terminology validator removed. |
| v2.19.0 | 2026-05-23 | Pre-promotion hardening: no behavior changes (catalog 63), a trust pass before active promotion. New cross-reference validator, .mdx count scanning, same-page anchor validation (23 pre-existing breaks fixed), an enforced version badge, pinned line endings, an honestly-scoped local validator bundle, and a branded 404 page. |
| v2.18.0 | 2026-05-21 | Highest-consensus PM skill gaps: 4 new phase skills (market-sizing, prioritization-framework, journey-map, survey-analysis), each refusing to fabricate data and labeling confidence honestly. Catalog grows 59 to 63; each ships with 3 thread samples. |
| v2.17.0 | 2026-05-20 | Native Claude Code sub-agent registration: the 4 sub-agent definitions moved to the canonical agents/ directory for @-mention auto-discovery; coordination dir renamed AGENTS/ to _agent-context/; frontmatter migrated to the metadata-nested structure; CI validators made bash-3.2 portable. 59-skill catalog unchanged. |
| v2.16.0 | 2026-05-17 | Active Orchestration + doc-stack modernization: 4 plugin sub-agents (pm-critic, pm-skill-auditor, pm-changelog-curator, pm-release-conductor) codify the Phase 0 adversarial-review loop and the 6-gate release runbook; 4 dispatch skills deliver cross-client access (validated on Codex CLI). Astro 5.13.x to 6.3.x + Starlight 0.39.x + Node 22.12+ across 5 CI workflows. Content catalog 55 to 59. |
| v2.15.2 | 2026-05-17 | v2.15.x Cycle Closeout + v2.16.0 Plan Reconciliation. Closeout patch successor to v2.15.1. No source-code, validator behavior, or catalog changes (55 skills + 27 enforcing validators unchanged). Planning-doc hygiene only: v2.15.x audit doc flipped from DRAFT to REMEDIATION SHIPPED with finding-by-finding closure table; v2.15.0 master plan “What’s next” updated; v2.16.0 repo-hygiene-plan + ci-plan + master plan reconciled against v2.15.1 shipped state (CONTEXT.md prereq marked DONE; net-new v2.16.0 validator scope reduced from “5 new” to “2 new + 1 extension” via carry-in reconciliation section); AGENTS/claude/CONTEXT.md refreshed to v2.15.2 SHIPPED state; version surfaces bumped (plugin.json, marketplace.json, README badge, .claude/pm-skills-for-claude.md); issue #132 [M-20] received v2.15.1 partial-close evidence comment |
| v2.15.1 | 2026-05-16 | Post-Tag Audit Remediation + Preventive CI. Same-day patch successor to v2.15.0. 55-skill catalog unchanged. Closes 18 audit findings: docs-site homepage refreshed (Tool card added; “40 skills” claims swept to 55); skills landing page Tool row added; workflow-generator silent-drop bug fixed (hardcoded workflow_info dict missing entries for 3 new sprint workflows); AGENTS.md command table extended with 23 missing rows (15 tool + 3 workflow + 5 pre-v2.15 legacy drift); sample-library README refreshed to 171 samples across 55 skills; Sprint Planning gets (agile) qualifier per new naming-discipline rule; new docs/concepts/sprint-skills-overview.md cross-family entry point; AGENTS/claude/CONTEXT.md refreshed (closes v2.14.x cleanup Task 2). 4 new preventive CI validators wired enforcing: check-landing-page-counts --strict, check-workflow-generator-coverage, check-agents-md-command-sync, plus pre-tag-validate orchestration script codifying the feedback_pre-tag-validator-bundle memory rule |
| v2.15.0 | 2026-05-16 | Sprint Skills Launch: 15 new skills under the new classification: tool taxonomy implementing Knapp/Zeratsky/Kowitz canonical sprint methodologies (7 Foundation Sprint family + 7 Design Sprint family + 1 tool-note-and-vote standalone). Skill catalog 40 to 55. Two family contracts with enforcing CI validators (each with —strict release mode); previously-undetected CI gap closed where neither family-specific validator was wired in. 3 new workflows including end-to-end FS-to-DS arc with 12-row slot-mapping table + 3-question go/no-go checkpoint (load-bearing replacement for dropped bridge skill). 45 library samples across 3 narrative threads (Brainshelf, Storevine, Workbench). Two user guides + two concept docs. Two adversarial-review cycles documented (FS-track 35 findings; DS-track 13 Codex + 20 self-audit findings; all P1 closed pre-tag) |
| v2.14.2 | 2026-05-10 | Codex final review closure (cumulative docs hygiene patch). 0 P0, 1 P1, 11 P2, 1 P3 findings addressed. validate-docs-frontmatter scope expanded to .mdx (parity with V6); check-no-body-h1 doc clarified with “what is NOT caught” framing; validate-mcp-sync guide refreshed for observe-mode default; sync-agents-md.yml workflow_dispatch hardened with two-layer defense (input gate + token gate); pm-skills-mcp README cross-repo update (5 “25 skills” residues to “40 skills”; v2.9.3 latest pointer; changelog table extended); CONTRIBUTING.md workaround count corrected; release plan + Release_v2.14.0 deferral table reframed. 40-skill catalog unchanged |
| v2.14.1 | 2026-05-10 | Polish + V15 regression fix: same-day patch for v2.14.0. Title duplication fixed across 62 hand-authored docs + 6 generator emission sites (Starlight title-vs-body-H1 migration regression); generator output reframed (63 caution-aside pages cleaned of contributor-noise); 182 en-dashes swept across 45 library samples; Mermaid 3-layer beautification + canonical style guide; two validators promoted to truly enforcing (check-internal-link-validity + validate-docs-frontmatter); new check-no-body-h1 validator added; MCP maintenance posture codified cross-repo |
| v2.14.0 | 2026-05-10 | Doc Stack Migration: MkDocs Material to Astro Starlight: doc-stack migration release with zero new skills (40 unchanged); 4 phases / 13 workstreams; production cutover via GitHub Pages source flip to Actions; Starlight-native asides + MDX details replace pymdownx admonitions; client-side Mermaid via astro-mermaid 2.0.1 with autoTheme; 115 library samples mounted under /samples/; 12 redirect entries preserved with /pm-skills/ base path; W13 B2.5 routing fixes (/reference/, /samples/, redirect base-path); Phase 0 Adversarial Review Loop via codex:rescue against trunk |
| v2.13.0 | 2026-05-05 | Foundation Hardening + Doc Stack Decision: maintenance and quality release with zero new skills (40 unchanged); 7 new CI gates that catch doc drift on PRs (validator inventory 15 to 22; enforcing tier 5 to 10); Diataxis-aligned doc structure with pm-skill-* filename prefix; Pattern 5C generated-content marker on 63 pages; out-of-cycle pm-skills-mcp v2.9.3 security patch (8 to 0 Dependabot alerts); Phase 0 Adversarial Review Loop applied (5 Codex review rounds + 3 resolution passes converged) |
| v2.12.0 | 2026-05-01 | OKR Skills Launch: foundation-okr-writer + measure-okr-grader pair covering the full quarterly write-and-score cycle; 38 to 40 skills; 6 new thread-aligned samples; Phase 0 Adversarial Review Loop applied |
| v2.11.1 | 2026-04-22 | skills.sh CLI compatibility patch; unblocks npx skills add product-on-purpose/pm-skills; lint hardening; em-dash sweep |
| v2.11.0 | 2026-04-18 | Meeting Skills Family (5 foundation skills + contract + enforcing CI) + lean-canvas; first cross-cutting skill-family pattern; 32→38 skills |
| v2.10.0 | 2026-04-11 | 3 utility skills: mermaid diagrams, slideshow creator, self-updater |
| v2.9.1 | 2026-04-10 | Workflows guide, count consistency CI, script docs enforcement |
| v2.9.0 | 2026-04-06 | Workflows: rename bundles → workflows + expand 3 → 9 |
| v2.8.2 | 2026-04-04 | Docs site polish + versioning concepts |
| v2.8.1 | 2026-04-04 | MkDocs Material documentation site launch |
| v2.8.0 | 2026-04-03 | PM skill lifecycle: Create, Validate, Iterate |
| v2.7.0 | 2026-03-22 | PM Skill Builder, enhanced CI, acceptance criteria |
| v2.6.1 | 2026-03-04 | Sample library recovery |
| v2.6.0 | 2026-03-04 | Claude plugin packaging |
| v2.5.2 | 2026-03-04 | Public doc hygiene |
| v2.5.1 | 2026-03-04 | Agent workspace canonicalization |
| v2.5.0 | 2026-03-02 | Foundation persona skill |
| v2.4.3 | 2026-02-16 | Release metadata |
| v2.4.2 | 2026-02-16 | Delivery plan policy |
| v2.4.1 | 2026-02-16 | Release doc consistency |
| v2.4.0 | 2026-02-16 | Contract lock and governance |
| v2.3.0 | 2026-02-14 | MCP sync blocking mode |
| v2.2.0 | 2026-02-13 | Guardrails and governance |
| v2.0.0 | 2026-01-26 | Flat structure restructure |
See the full Changelog for a condensed view.