メニュー
Contribution loop for agents working in a Cogni node-template repo. Use when starting implementation, preparing a PR, or checking node/operator ownership boundaries.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when adding scheduled, recurring, or AI workflow behavior to a Cogni node; deciding graph vs route vs own Temporal worker; working with /api/v1/schedules, GraphRunWorkflow, NodeTaskWorkflow, defineScheduledJob, or the shared scheduler-worker substrate.
Use whenever you customize a Cogni node's visual identity — branding, icon, name, theme colors, metadata, the public homepage, or chat suggestions. Triggers on "style this node", "rebrand", "customize the node", "make the homepage for <mission>", node launch/formation styling. The homepage is the main event, not an afterthought.
Orchestrate multiple dev agents (spawned subagents OR human-driven worktrees) against ONE story-level outcome — hold the e2e vision, decompose into non-overlapping linked tasks with freeze/secrets guardrails baked in, monitor the tasks for movement, and intervene only on collision or drift. Use when a problem is bigger than one PR and needs 2+ agents working linked tasks under a shared contract (node-template feature builds, substrate ports). Triggers: "manage these devs", "coordinate subagents", "split this into tasks for N agents", "hold the story while agents work the tasks", "dev manager".
Authoritative reference for THIS node's test setup — the test layers it actually ships (unit, component, external, stack), the vitest configs, what runs in CI vs locally, testcontainers/RLS component setup, fake adapters + APP_ENV=test, and the non-obvious gotchas (dotenv CWD trap, skip-gate ordering, never-mock-the-DB). Use whenever writing a new test, deciding "which layer does this belong in", debugging a flaky test or a `.env.test`-not-loading error, running any `pnpm test:*`, working with testcontainers, touching fake adapters, or wondering whether a test runs in CI. This node is a single flat-layout app — not the operator monorepo.
Close the deploy_verified loop for THIS node after a PR is flighted to its candidate deploy. Confirm the candidate build matches the PR head SHA via /version, enumerate impacted surfaces (API routes, UI pages, graphs), exercise each against the real deployed URL on both the human (Playwright) and agent (API) axes, query Loki for your own request when creds are available, then post an approve/fail scorecard as a PR comment. Use whenever the user says "validate the candidate deploy", "prove this PR on candidate", "close the deploy_verified loop", or runs "/validate-candidate" (with or without a PR number). NOT for pre-merge CI or local dev testing — this runs after the operator has flighted the PR.
Use any time you are about to add, change, or migrate a Postgres or Doltgres table in this node — editing schema TS under `packages/db-schema/src` or `packages/doltgres-schema/src`, running `drizzle-kit generate`, writing a migration `.sql`, touching `meta/_journal.json` / `*_snapshot.json`, adding row-level security / tenant isolation (a table with a FK to `users` or `billing_accounts`), or debugging a migration that didn't apply on candidate. Mandatory before any schema edit. Covers the RLS-first rule + the component gate that enforces it.
| name | contribute-to-cogni |
| description | Contribution loop for agents working in a Cogni node-template repo. Use when starting implementation, preparing a PR, or checking node/operator ownership boundaries. |
Read docs/guides/contributing-to-cogni.md and root AGENTS.md.
Core contract:
.cogni/secrets-catalog.yaml; never write secret values into git.pnpm check before publishing.contribute-knowledge-to-cogni.