Second Brain

Wire all installed plugin primitives into Helix Layer 1-7 + Meta + Ses…

2 min read

Wire all installed plugin primitives into Helix Layer 1-7 + Meta + Ses…

Decision

Wire all installed plugin primitives into Helix Layer 1-7 + Meta + Session-close + Documentation-deliverables stages; codify canonical pick for 9 overlapping capabilities (skill-creation→superpowers:writing-skills, recurring-loops→loop, review→pr-review-toolkit:review-pr, etc.); install nightly plugin-version-drift detector; upgrade plugin-integration.md as single source of truth for plugin × workflow × Helix-layer mapping.

Rationale

Audit triggered by AJ’s question after 25-Apr Superpowers self-update incident broke brainstorm.service. Pre-session state: ~55% plugin utilization (Superpowers core mandated, ~7 other plugins under-used or dormant). Three structural defects discovered: (a) no Helix-layer × plugin-primitive matrix → orchestrators reinvent what plugins provide (Law 10 violation); (b) triple/double overlaps (skill-creator×3, excalidraw×2, schedule×3) without canonical pick → behavioral drift across sessions (Law 2 violation); (c) no drift detector for plugin auto-upgrades → 25-Apr incident pattern can recur (Law 12+13 violation). Fix landed in 6 cascade-linked memory files (governed by setup-curator) + 1 cron at /root/.claude/crons/plugin-version-drift-check.sh (02:30 IST nightly, 30-day snapshot retention, Telegram-on-drift, DRY_RUN-tested clean+drift paths). 8/8 curator verification checkpoints GREEN. detect_biases scan returned 0 biases on the recommendation reasoning. Three strategic decisions surfaced for AJ override: auto-update-on policy, superpowers:writing-skills as canonical, layered Layer-6 ownership (quality-orchestrator + pr-review-toolkit:review-pr) — implemented per recommendations under feedback_autonomous_by_default.md while flagging for explicit override. assess_reasoning_quality scored 25% (low) — honest reflection that the audit was tooling-inventory-driven rather than pure-cognitive-ST: 2 thoughts, 0 hypotheses, audit rigor lived in the 8-point curator checklist + DRY_RUN cron tests rather than ST primitives. The real evidence chain is the curator verification, not the ST chain.

Alternatives Rejected

Outcome

Pending

Graph View

Backlinks