From second-brain
Manual deep-dive on the current/most-recent session. Proposes up to 3 grounded "pin" candidates the user can accept/reject/edit. No autonomous critic.
npx claudepluginhub cain-ish/claude-code-plugin --plugin second-brainThis skill is limited to using the following tools:
Generate up to 3 candidate "pins" from this session and let the user accept/reject/edit each.
Mandates invoking relevant skills via tools before any response in coding sessions. Covers access, priorities, and adaptations for Claude Code, Copilot CLI, Gemini CLI.
Share bugs, ideas, or general feedback.
Generate up to 3 candidate "pins" from this session and let the user accept/reject/edit each.
Read session activity since SessionStart. Identify:
Read ~/.second-brain/persona-signals.jsonl (if it exists). Find entries where:
count >= 3 (pattern observed across 3+ sessions)graduated == falseFor each qualifying signal, create a candidate pin:
signal field, rewritten as a concise USER.md rulecode_style → ## Approach or ## Hard Rulescommunication → ## Communicationworkflow → ## Approachtooling → ## Hard Rulesdecision_making → ## Approachcount and top 2 evidence entries so the user can verify the pattern is realAfter user accepts: call pin_to_user, then edit persona-signals.jsonl to set graduated: true on that entry so it won't be re-proposed.
For each candidate, ask: "Would the user want to remember this in 2 weeks?" If no, drop. Keep at most 3 candidates.
For each surviving candidate, show:
For each accepted candidate, call the appropriate MCP tool: pin_to_user, pin_to_project, or archive_to_wiki. For wiki entries that don't fit the pin tools, write directly via Write tool.
Report what was pinned. No critic-log, no decay tracking, no auto-extraction.