PM Release Conductor (Dispatch Skill)

PM Release Conductor (Dispatch Skill)

Cross-client dispatch wrapper for the pm-release-conductor sub-agent. Detects runtime; dispatches to the native sub-agent on Claude Code; reads agents/pm-release-conductor.md and inlines chain composition on non-Claude clients via "reference + execute inline" pattern.

Status summary (v2.16.0): PRODUCTION on Claude Code (native sub-agent path). DRY-RUN VALIDATED on Codex CLI 2026-05-17 per gate-test-results_2026-05-17_codex.md; LIVE release on Codex CLI is NOT independently exercised, so use with caution and run --dry-run first as a rehearsal. EXPERIMENTAL on Cursor / Windsurf / Copilot CLI / Gemini CLI (UNTESTED at v2.16.0 ship).

See Sub-Agent Compatibility Matrix for the canonical safe-usage matrix + what-was-validated detail + v2.17 expansion plan. For live release on a non-Claude client, ideally re-run the harness at maintainer-gate-testing-codex.md on that specific client first.

When to Use

• You are running a pm-skills release on a non-Claude client (Codex CLI, Cursor, Windsurf, Copilot, Gemini CLI)
• You want the same 6-gate discipline that pm-release-conductor enforces on Claude Code, with auditor + curator behaviors inlined at the relevant gates
• You explicitly want skill-invocation semantics over sub-agent semantics on Claude Code (rare; the native sub-agent is preferred on Claude Code)

When NOT to Use

• You only need to review a PM artifact -> use utility-pm-critic
• You only need a governance audit (not a release) -> use utility-pm-skill-auditor
• You only need a CHANGELOG draft (not a release) -> use utility-pm-changelog-curator
• You want to perform release operations WITHOUT explicit gate confirmation -> the conductor refuses bypass; manual release outside the conductor is the right path

Instructions

Runtime detection step. Determine which AI client is invoking this skill.

If you are running in Claude Code with the pm-skills plugin installed

Invoke @agent-pm-skills:pm-release-conductor with the user's target version + optional flags from $ARGUMENTS. The native sub-agent walks the 6 gates and natively chains to pm-skill-auditor (G0, G2.5) and pm-changelog-curator (G2) via the Agent tool. Return the conductor's gate-by-gate output to the user.

If you are running in any other AI client

Codex CLI, Cursor, Windsurf, Copilot, Gemini CLI, or any other client without native pm-skills plugin sub-agent support:

1. Read the canonical sub-agent definition at agents/pm-release-conductor.md
2. Read the canonical runbook at docs/contributing/release-runbook.md (the conductor's referential source for gate definitions)
3. Execute the system prompt body as your operating instructions
4. Walk the 6 gates inline. At gates that require chain composition:
   • G0 (Pre-tag readiness): instead of chaining to pm-skill-auditor, read agents/pm-skill-auditor.md and execute the auditor's 4-step audit flow inline. Capture the layered output (full findings + Status Summary + Status YAML). Treat the Status YAML as your G0 sub-check 5 input.
   • G2 (Version bump + CHANGELOG prep): instead of chaining to pm-changelog-curator, read agents/pm-changelog-curator.md and execute the curator's 8-step drafting flow inline. Capture the layered output. Treat the Status YAML as your G2 sub-check 3 input.