Menu
Use when the task needs validation, testing, completion checks, or evidence review.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Use as the canonical first-stage Agmo workflow when the user wants idea exploration, design tradeoffs, or requirement shaping before planning or implementation.
Run a Codex-native structured code review using agmo-critic and agmo-verifier where evidence is needed.
Harvest `debt:` markers across the repo into a read-only ledger without modifying code.
Audit and maintain the vault-backed llm-wiki knowledge base, including stale pages, low-confidence or contested captures, superseded captures, missing frontmatter, and broken wikilinks.
Use when prior notes, durable project knowledge, or vault-backed context should influence the current task.
Use when one execution lane is no longer enough and work should escalate to the durable `agmo team ...` runtime.
| name | verify |
| description | Use when the task needs validation, testing, completion checks, or evidence review. |
| argument-hint | [artifact, change, or implementation to verify] |
Use this when the primary goal is validation rather than planning or first-pass implementation.
The main session orchestrates verification and delegates the evidence-gathering lane to agmo-verifier.
Typical responsibilities:
executeWhen spawning native subagents for this workflow, keep each agent id until its result is integrated, then call close_agent for completed, failed, superseded, or no-longer-needed lanes so thread slots are released before the next delegation.
plan-review)brainstorming)Return a concise PASS / FAIL / INCOMPLETE / BLOCKED verdict plus gaps.
FAIL means the proof shows broken behaviorINCOMPLETE means the proof bar was not fully met yetBLOCKED means the next verification step cannot proceed without external resolutionDistinguish missing proof from proven failure so execute can decide whether to fix, re-run, or report a blocker.
Before ending a meaningful verification stage, make the final response or delegated result save-ready:
execute, plan, or done