orchestrator

Browse the orchestrator's knowledge graph - search, filter by type/domain, inspect notes, resolve threads

First-run onboarding - bootstrap the orchestrator's knowledge from your project

Trigger orchestrator knowledge maintenance: consolidation, confidence decay, gap analysis

Show current orchestrator state: open threads, checkpoint, neglected areas, autonomy scores

Persistent memory concierge that curates knowledge retrieval for the calling agent. Tracks what has been communicated, detects compaction, and strategically mixes fresh and refreshed context. Invoke with model sonnet for routine, opus for complex judgment.

Autonomous knowledge maintenance agent - consolidation, decay, gap analysis, revalidation

Use when an open question, investigation, or tracked issue has been fully resolved. Prevents future sessions from revisiting solved problems and keeps the briefing focused on what's actually still open.

Use instead of direct lookup when you need curated, context-aware knowledge retrieval - concierge searches, synthesizes, and won't repeat what it already told you.

MANDATORY every turn. Evaluate which orchestrator tools and skills apply to what you're about to do, what just happened, and what you learned. This is your orchestrator intercept - run it before and after acting.

Use proactively when you discover a bug, known issue, footgun, security concern, or limitation that future sessions should be aware of. Also use when a debugging session reveals a non-obvious root cause or when you find something that could silently break.

Use when beginning any task, switching to an unfamiliar area of the codebase, or when context from previous sessions would help. Also use when resuming after context compaction.

Use proactively whenever you discover a pattern, convention, gotcha, workaround, or important fact about the codebase that would be valuable in future sessions. Also use when the user corrects you or explains how something works. If in doubt about whether to record it, record it.

Use proactively after making or recommending an architectural decision, choosing between implementation approaches, or establishing a new convention. Captures both the decision AND the reasoning so future sessions understand the 'why' and don't revisit settled questions.

Always active. You are an orchestrator first - scan your full toolkit every turn, use what fits, capture everything that matters.

Use when facing a complex task that would benefit from checking past decisions, known constraints, or previous failed approaches before diving into implementation. Gathers domain-specific context to avoid contradicting established patterns.

Use after a debugging session reveals root causes, when an approach failed and you pivoted, or when an unexpected issue occurred. Captures what happened and what to do differently so the same mistake isn't repeated across sessions.

Use when the user expresses a preference about workflow, coding style, communication, tooling, or process. Examples: 'always use bun', 'don't auto-commit', 'I prefer X over Y', 'never do X without asking'. These preferences persist across sessions and projects.

Use when you encounter a design choice, naming convention, architectural pattern, or implementation approach and wonder why it was done that way. Use before proposing changes to check for prior decisions. Use when you're about to make an architectural choice and want to avoid contradicting past work.

Use when finishing a task, completing a significant milestone, or when the conversation is ending. Saves progress so the next session picks up seamlessly.

5 hooks across 5 events