plugin

Use when work is multi-step, long-running, side-effecting, approval-gated, or must survive a coding-host session and remain observable through APXM.

Use when designing APXM-owned event-triggered loop specs or workflow packs, and when identifying which APXM server/MCP/Dekk capabilities are missing. Do not use as the primary skill for generic MCP usage, workflow following, worker selection, or execution.

Use when converting a Codex skill, Claude/Superpowers-style skill, or local agent procedure into an APXM workflow pack with AIR source, policy, compile targets, verification, and conversion notes.

Use when a decision needs a traceable APXM council with multiple worker perspectives, dissent, critique, evidence, and a final recommendation. Use for architecture choices, risk reviews, plan reviews, and high-impact tradeoff analysis.

Use when following, watching, replaying, archiving, or explaining APXM workflow/run progress through APXM OS observability surfaces such as watch, rollout list, rollout replay, rollout archive, process list, and workflow sessions.

Use when validating, compiling, executing, running, analyzing, or explaining APXM workflow artifacts such as canonical .air workflows, Python frontend workflows, .apxmw workflow files, and compiled APXM objects.

Use when verifying APXM workflow traces, artifacts, schema validity, test evidence, worker outputs, budget records, or whether an APXM run actually satisfied its declared acceptance criteria.

Use when discovering, validating, selecting, or briefing APXM workers, including asking a verified worker to propose an APXM workflow. Use whenever orchestration depends on which agents or LLM hosts are registered and ready.