richfrem/agent-plugins-skills / plugins/agent-agentic-os/.apm/skills/running-control-plane
running-control-plane
FlaggedCRITICAL INTAKE GATEWAY: use at the start of any non-trivial engineering task, feature, refactor, multi-file bugfix, work package or new idea, including requests to explore, brainstorm, build, fix or change something, before planning or code changes. Enforces read-only discovery, 1-3 structured questions with defaults, control-plane registration and the immutable four-pillar TASK_SPEC.md. Also runs the rest of the control plane: health ("run health check", "check os metrics", "system monitor", event bus, locks, memory, "is my signing identity ready" status check), transition simulation ("dry run transition", "simulate transition", "check edge X -> Y"), worktree selection, and gating on/off ("disable the control plane", "turn off the push guard", "stop gating my commits"). Not for project setup or key creation (use setting-up-agentic-os) or improving skills (use improving-skills-in-lab).
references/multi-round-external-review-protocol.md
View on GitHub# Multi-Round External Review Protocol
Applies at both `MULTI_AGENT_REVIEW` (plan) and `MULTI_AGENT_CODE_REVIEW` (implementation) gates
whenever the user opts into external review rather than skipping it. Builds on
`bundling-context`'s generic Multi-Persona Fan-Out mechanics (per-persona `responses/` subfolder,
regenerate-in-place per round) with the review-specific judgment calls those generic mechanics
deliberately don't encode.
## Persona selection is content-driven, not fixed
The Graph Planning Phase 1 Fan-Out Trio (Architecture Skeptic, Security/Edge-Case Auditor, TDD
Contract Reviewer) is a default, not a mandate. Add or drop personas based on what the round
actually touched:
- A round that added significant new test coverage warrants `tdd-contract-reviewer.md` even if
it wasn't in an earlier round.
- A round that's purely cosmetic/docs may warrant only one persona, or skip this gate entirely
(see Path B in `SKILL.md`).
- A round touching security-sensitive logic (auth, injection surfaces, gate enforcement) always
warrants `adversarial-security-auditor.md`.
## Folding prior rounds' findings into the next round's brief
Before regenerating `prompt.md` for round N+1, update the leading context brief to state round
N's findings **in past tense as already-fixed** (or explicitly deferred, with why), not silently
dropped:
- List each finding with a `[FIXED]` or `[DEFERRED — reason]` tag and one line on what changed.
- Ask reviewers to verify the fix closed the gap, not rediscover the original finding — explicitly
say "don't re-litigate finding N unless something about *how* it was fixed looks wrong."
- Carry forward anything genuinely still open (e.g. a finding whose real fix needs its own design
pass) as an explicit "still open" item, not silently.
This keeps review rounds convergent — each round narrows toward genuinely new findings instead
of re-discovering the same ones.
## Plan review and implementation review are not interchangeable
Skipping `MULTI_AGENT_REVIEW` (plan) and only running `MULTI_AGENT_CODE_REVIEW` (implementation)
after the fact is not equivalent to running the plan gate — it's post-hoc auditing of already-
built work, which is more expensive to act on than catching a design flaw before implementation.
Both gates existing independently in the state machine is deliberate; treat a skip of one as a
real decision (recorded via `record_review_skip`, per `agent_control.py`'s `GATE_REQUIREMENTS`),
not a default.
## Reference
See `plugins/dev-utils/skills/bundling-context/SKILL.md`'s Multi-Persona Fan-Out Mode section for
the underlying generic folder/regeneration mechanics this protocol builds on.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |