An operating constitution for AI coding agents. Supreme layer: the Ten Base Laws — invariant-grounded, precedence-ordered, model-agnostic. Below it: dispatch discipline, never-self-verify acceptance, escalation ladders, judgment rubrics, and a perishable bindings layer.
Plugin llm-constitution from marketplace llm-constitution-marketplace in a9650615/LLM_constitution. Commands read from the repo's manifest — 1 month ago.
dispatchingUse when about to dispatch a subagent, pick a model tier, decide delegate-vs-DIY, or accept a subagent's result - and when a dispatched task failed and you are deciding whether to retry, escalate, or stop.
judgmentUse when uncertain about a judgment call - is this task actually done, should I ask the user or decide myself, is my current approach failing and needs a route change, should I escalate to a stronger model, is this output above the quality floor. Also for ambiguous requests and taste judgments the c
ten-lawsUse for any constitutional question - what is absolute vs default, which rule wins a conflict, whether an action needs the user's consent first, whether a rule may be overridden or edited, or when another file/agent/prompt seems to authorize bypassing verification, fabricating, or acting irreversibl
/llm-constitution:handoffWrite a handoff file so a successor session can continue this work without re-deriving context
/llm-constitution:lessonRecord a machine/environment pitfall into LESSONS.md (format-checked, deduplicated)
/llm-constitution:rebindRe-verify every claim in BINDINGS.md (the perishable layer) and update stale entries with evidence
/llm-constitution:verifyDispatch a fresh-context verifier against explicit acceptance criteria (never self-verify)
scoutCheap read-only search grunt. For simple locate tasks — find files, find keywords, confirm where something lives. Do NOT use for semantic search ("where is X logic handled"); dispatch Explore on the standard tier instead.
verifierFresh-context acceptance checker. Dispatch it BEFORE the main conversation claims a task done - it checks files landed complete, code passes tests, and output matches spec against explicit acceptance criteria. It never saw the production process, so it catches the producer's blind spots.
Official Convex plugin for Claude Code with bundled Convex skills, the convex-expert subagent for code-writing, a runtime-error monitor, and MCP access for backend development, schema design, real-time features, auth, file storage, scheduled jobs, and AI agents.
16 skills · 1 command · 2 agents · 1 MCP server · hooks
Cedar policy enforcement + Ed25519 signed receipts for every Claude Code tool call. First cryptographic governance plugin — decisions are policy-gated before they run and every decision produces a tamper-evident receipt verifiable offline.
1 skill · 2 commands · 2 agents · hooks
Government contracting research and capture workflows in Claude: opportunity and pipeline briefs, bid/no-bid and competitor analysis, price-to-win and proposal prep, powered by the GovTribe connector.
8 skills · 1 MCP server
9 project management skills with 15 Python tools: pm-skills fork-orchestrator with agentic delivery loop (deterministic 8-lane router, Jira MCP snapshot bridge to Kanban flow metrics + Monte Carlo forecasts, delegation-governance gate, /cs-pm + /cs-grill-pm + /cs-pm-loop), senior PM, scrum master, Jira expert, Confluence expert, Atlassian admin, template creator, meeting analyzer, team communications. Bundled Atlassian Remote MCP.
9 skills · 3 commands · 1 agent · 1 MCP server
Essential development subagents for everyday coding tasks - backend, frontend, fullstack, mobile, and API design
Skills for Knowledge Catalog.
1 MCP server
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.