Menu
Run the 6-question phase-end checklist when a logical phase completes (after a feature, before a context switch, or when the Stop hook prompts you). Use to surface skipped Iron Rules, missed memory writes, and overdue /compact opportunities.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Walk through stale or conflicting memory entries and routing-table drift. Use when the user runs /sustain:audit, asks "is my memory store healthy?", or before a major release of personal rules.
Search persistent memory across mempalace MCP, claude-mem MCP, and the structured filesystem fallback. Use when the user asks "how did we solve X before?", "what did we decide about Y?", or anything that implies recalling prior-conversation context.
Decide whether and where to persist a learning. Use at phase boundaries when something seems worth long-term memory — a rule, a non-obvious decision, an external resource, or a project-state change. Picks the right wing (feedback / project / reference / user) and routes through the user's active memory backend.
Interpret session token statistics and recommend the next move — /compact, delegate to a subagent, switch to a smaller model, or stop and persist learnings. Use after `/sustain:status`, when the session feels long, or when context-window pressure is showing up in degraded responses.
Establish Iron Rules (3+1) and R1–R5 token-saving rules at session start or whenever the rules feel forgotten. Use when user asks about token efficiency, when rules need re-grounding, or proactively at the start of complex multi-turn work.
| name | phase-self-check |
| description | Run the 6-question phase-end checklist when a logical phase completes (after a feature, before a context switch, or when the Stop hook prompts you). Use to surface skipped Iron Rules, missed memory writes, and overdue /compact opportunities. |
Walk through these six questions. For each, answer "yes/no/n-a" and act on the "yes" answers before moving on.
Iron-1 trigger: Did the user's most recent intent contain "逐個 / 仔細 / 全部 / 不要省略 / thoroughly / each / all / every", or was the task security-related? If yes → for the rest of this work, no shortcuts.
Subagent hygiene (Iron-2): Did you call a subagent in this phase, or are you about to? If yes → confirm the prompt has both: (a) a word cap appropriate to the depth (≤200 / ≤800 / ≤2000) and (b) an escape clause "but report anomalies in full regardless of cap".
Repeated edits (Iron-3): Did you make the same string edit twice or more? If yes → switch to perl -i -pe 's|old|new|g' files... and run git diff.
Templating (Habit): Did you plan N similar tasks? If yes → look for the largest common factor; if templatable, validate on one sample before bulk producing.
Context size: Is the conversation context noticeably large or are you crossing a phase boundary? If yes → consider /compact to drop early raw output.
Memory candidate: Is there an observation from this phase worth long-term memory — a rule learned, a decision rationale, an external resource pointer, a project-state change? If yes → write it as a feedback / project / reference memory, with a "why" line so future sessions can judge if it's still load-bearing.
/skill phase-self-check.