refactor: pm-stack → Portfolio Manager bundle (manager-of-teams tier)

[mabry1985]

What

Completes the portfolio→team split (ADR 0055). The Lead Engineer bundle is the team tier (project_board + a repo's coding agents); pm-stack is now the manager tier that orchestrates and spawns those teams.

Changes

• enabled: [delegates, portfolio] — the manager spine only (was: also project_board + agent_browser). The PM runs no board of its own.
• project_board + agent_browser stay installed but off. A spawned team's plugins.dir defaults to this host's plugins dir, so the teams the PM spins up discover them here without a per-team reinstall. Install once, provision many teams.
• portfolio pinned v0.5.0 → v0.7.2 — the team-spawning release (portfolio_spinup_team / teardown_team / autodispose / teams).
• archetype: Project Manager → Portfolio Manager, with a starting soul (the spin-up → dispatch → roll-up → sequence → dispose loop).
• portfolio.team_template wired — empty falls back to the plugin's shipped example; point it at a creds-filled copy for teams that run real model turns.
• verified_against 0.35.3 → 0.66.0; README rewritten for the manager tier.

One-liner it enables

portfolio_spinup_team(name="docs-team", repo="/abs/path")   # boots + registers a board
portfolio_dispatch(board="docs-team", title=…, spec=…)
portfolio_autodispose()                                     # dispose when the board drains

🤖 Generated with Claude Code

[protoquinn]

👀 Quinn is reviewing — verdict (PASS / WARN / FAIL) + findings to follow.

[protoquinn]

QA Audit — PR #9 | refactor: pm-stack → Portfolio Manager bundle (manager-of-teams tier)

VERDICT: PASS

---

CI Status

• No CI checks found (Gap — repo may not have CI configured)

Diff Review

• Clean manifest refactor: enabled narrowed to manager tier [delegates, portfolio]; team plugins (project_board, agent_browser) stay installed-but-disabled for spawned-team discovery — consistent with ADR 0055
• portfolio pin v0.4.0 → v0.7.2; verified_against 0.35.3 → 0.66.0; archetype relabeled as Portfolio Manager with soul
• README fully rewritten for the manager-of-teams tier with spin-up/dispatch/autodispose examples — clear and accurate

Observations

• GAP: No CI checks — verified_against: 0.66.0 claim cannot be confirmed; verify manually before relying on this bundle
• GAP: portfolio-plugin tag v0.7.2 unverified (tooling can't reach external repos) — confirm the tag exists and carries the claimed portfolio_spinup_team / autodispose tools
• GAP: Clawpatch structural review unavailable (pm-stack not in project registry) — no cross-file analysis beyond diff inspection
• LOW: PR claims "Generated with Claude Code" — manifest refactor is low-risk and well-documented, but human sign-off on the verified_against bump is advisable

— Quinn, QA Engineer

[protoquinn]

Submitted APPROVE review on protoLabsAI/pm-stack#9.

[protoquinn]

QA Audit — PR #9 | refactor: pm-stack → Portfolio Manager bundle (manager-of-teams tier)

VERDICT: WARN

---

CI Status

• verify: queued

Diff Review

• Pure manifest + docs refactor (2 files, 217 lines); no executable code changes.
• protoagent.bundle.yaml: enabled narrowed to [delegates, portfolio] — manager spine only. project_board + agent_browser stay installed-but-off as a discovery layer for spawned teams. Rationale documented inline.
• portfolio pinned v0.5.0 → v0.7.2 (team-spawning release); verified_against 0.35.3 → 0.66.0. README rewritten for the manager-tier narrative with usage examples.
• Archetype updated: label, icon, blurb, and a new soul persona.

Observations

• LOW: Clawpatch structural review unavailable — protoLabsAI/pm-stack not in the project registry. Proceeding with diff-level review only.
• LOW: verified_against jumps from 0.35.3 to 0.66.0 (~30 minor versions). The verify CI check is still queued — no smoke-test confirmation yet that this pin set actually works against protoAgent 0.66.0. Once CI passes, this concern evaporates.
• LOW: portfolio.team_template defaults to empty string, which falls back to the plugin's shipped example. This is a sensible default, but the comment suggests pointing it at a creds-filled copy for real model turns — worth calling out in the README's Install section (already done).

No blocking issues. The design is consistent with ADR 0055 and the documentation is clear. Awaiting CI completion for final PASS/FAIL.

— Quinn, QA Engineer

[protoquinn]

Submitted COMMENT review on #9.