메뉴
Use when adding a new ingest stage or a new SurrealDB table to the ax graph. Encodes the exact registration gotchas that fail CI if missed.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Guided experiment-loop retrospective over the ax agent-experience graph. Walks the user through their open proposals (accept-with-scaffold or reject), pending verdicts (confirm the suggested verdict or override), and recent harness-hook effectiveness signal. Triggers when the user says "let's do an ax retro", "ax retrospective", "review my ax proposals", "triage proposals", "experiment loop status", "lock pending verdicts", "hook effectiveness review", "intervention review", "self-improvement session", or invokes /ax:retro. Reads/writes via the local `ax improve` and `ax hooks` CLIs. Do NOT auto-trigger on unrelated work.
Definition-of-done checklist for shipping a new write, signal, table, edge, or query to the ax graph. Use when adding a SurrealDB table/edge/field, an ingest derive-stage, a new analytic query, or any new ax capability - before opening the PR. Ensures every write gets an on-demand read AND a proactive (agent-facing) read AND docs/distribution, not just the write. Triggers on "ship", "new signal", "new table/edge", "new lens/query", "wire this up", "is this done", or finishing an ax feature branch.
Model-routing orchestration for any expensive frontier model (Fable, Opus, GPT-5.x) - the main model keeps judgment and Q&A review, mechanical subagent dispatches carry an explicit cheaper model, and ax measures whether the routing actually worked. Use when orchestrating codebase-heavy or token-heavy work with subagents, when dispatching Agent tasks without a model, when the user says "route to cheaper models", "efficient dispatch", "optimize model spend", or asks where their token spend goes. Pairs with the route-dispatch hook (deterministic backstop) and `ax dispatches` (evidence). Do NOT fire on single-shot questions or tiny tasks with no dispatching.
Surplus-quota training loop over the ax graph - the agent burns the remaining 5h/7d plan-quota window on self-improvement: locking pending verdicts, filling briefs, backtesting routing classes, minting proposals, running worktree experiments, and drafting upstream issue reports. Triggers when the user says "/dojo", "enter the dojo", "dojo time", "train overnight", "burn my surplus quota", "dream mode" (legacy name), or invokes /loop /dojo. Requires ax (axctl) on PATH and the local SurrealDB running. Do NOT auto-trigger on unrelated work or when the user merely mentions quotas.
Install + verify ax (the agent experience layer). Triggers when the user says "install ax", "set up ax", "ax not found", "ax doctor", "is ax running", "fix ax install", "first-time ax setup", or any setup question about the ax CLI / skills / daemon. Walks the install via the install.sh + skills.sh + first ingest, validates with `ax doctor`, and points the user at ax:retro (experiment loop) and ax:extract-workflow (reconstruct workflow behind shipped artifacts).
Write the agent-generated narration of the current session - the reviewable story of what changed, including what never reaches a PR (user corrections, abandoned attempts, tool failures). Triggers on "narrate this session", "summarize what changed", "write the session story", "narrate what we did", "session narration". Output is .ax/narrations/<session-id>.json for the ax studio narration view. Do NOT fire on "summarize this file" or generic recap questions answered inline - this skill writes a structured artifact.
| name | ax-add-ingest-stage |
| description | Use when adding a new ingest stage or a new SurrealDB table to the ax graph. Encodes the exact registration gotchas that fail CI if missed. |
Hard-won checklist - each step below has a CI gate that fails if you skip it.
DEFINE TABLE ... SCHEMAFULL + fields to packages/schema/src/schema.surql (top-level fields explicit; nested objects → JSON-encoded string; datetimes via JS Date).SCHEMA_TABLES (apps/axctl/src/queries/insights.ts) - a mirror test diffs DEFINE TABLE names vs SCHEMA_TABLES and fails CI if missing.bun install so @ax/schema resolves to the worktree copy (not the main tree's symlink) before the mirror test will pass.derive-tagged ingest stageStageDef at the bottom of the stage file (mirror apps/axctl/src/ingest/derive-opportunities.ts): export const FooKey = Schema.Literal("foo") + fooStage.apps/axctl/src/ingest/stage/registry.ts: add the import, add FooKey to the IngestStageKey Schema.Union([...]), add fooStage to ALL_STAGES.apps/axctl/src/cli/effect-cli.test.ts: bump the resolveIngestStages: default runs every stage .toHaveLength(N) by 1, and add your key (sorted) to the --derive-only list. CI fails otherwise.Effect.catchCause returning a zero-row stat, so a stage error never aborts the surrounding ingest.Document it in BOTH cli-reference gates or CI fails: docs/cli.md (or README) + apps/site/public/llms.txt (scripts/check-cli-reference.ts), AND a card in apps/site/app/routes/docs/-cli-reference.data.ts (scripts/check-site-cli-reference.test.ts).
Schema.Literal(...) collapses → use Schema.Literals([...]).Schema.Date doesn't JSON round-trip → use Schema.DateFromString.check(Schema.isDateValid()).