메뉴
Use when an active, human-approved loen run should keep going multi-turn on its own — wraps it in Claude's native /goal from loop.yaml. Optional; never bootstraps a run or submits /goal itself.
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
| name | loop-goal |
| description | Use when an active, human-approved loen run should keep going multi-turn on its own — wraps it in Claude's native /goal from loop.yaml. Optional; never bootstraps a run or submits /goal itself. |
Wrap the ACTIVE loen run in a native /goal condition so "keep going until the
gates are green" runs multi-turn without hand-holding. Invoked as
/loen:loop-goal (optionally with an explicit run-id). This wrapper NEVER
weakens the loop protocol: loen:audit stages, the loop-guard hook, and the
human approval gate stay exactly as in the MVP — /goal only automates the
turn loop.
docs/loen/current resolves to a
run directory) whose contract was human-approved and whose loen:audit plan
returned OK (both recorded in docs/loen/<run-id>/state.md). No active
run → STOP and point to /loop-delivery, /loop-repair, or
/loop-autoresearch — this skill wraps an existing run, it never bootstraps
one. One goal run wraps ONE loen run: if the user passed a run-id that
differs from the active run, REFUSE (mirror of the hook's cross-topic
block).python3 <skill-base>/../../scripts/make_goal.py docs/loen/current/loop.yaml.
Exit 1 means the contract is not audit-plan shaped — report its stderr and
stop. Show the produced string to the human VERBATIM for them to submit as
/goal …. NEVER submit /goal yourself — it is a native user-level
command; the human stays in control of granting multi-turn autonomy./goal
mechanics: the /goal evaluator only reads the transcript — it runs no
commands. During the goal run the worker MUST print every gate command, its
exit code, and metric summaries into the conversation; a condition like
"all tests pass" without printed evidence never evaluates true. The
generated string already encodes this ("… prints each command's output
summary as evidence") — do not trim it./loop recipe (long-running gates). For gates that poll external state
(CI runs, deploys), offer: /loop <interval> loen:audit check with
<interval> matched to the external system's cadence. The durability
warning is MANDATORY: /loop is session-scoped — it dies with the session
and recurring tasks auto-expire; durable scheduling belongs to Routines /
OS schedulers / CI, outside this skill's scope.handoff_conditions keep
hard-stopping as usual.loop-delivery, loop-repair,
loop-autoresearch, or loen:audit references this skill.Use when improving ONE numeric metric under a controlled research loop with a fixed eval and kept/reverted experiments. Not for a feature (use loop-delivery) or a failing test (use loop-repair).
Use when delivering ONE bounded change — a feature, refactor, or chore — as a controlled, audited loop in any repo. Not for a failing test (use loop-repair) or a numeric metric (use loop-autoresearch).
Use when a specific test, CI job, or regression is failing and must be fixed under a reproduce-first controlled loop with proven regression coverage. Not for delivering a change (use loop-delivery) or metrics (use loop-autoresearch).
Use when a loen loop stage — plan, act, check, or result — must be validated and gated before the next one. Mode-aware for delivery/repair/research; the execution-loop analog of check-chain.
Use when you need a cross-run dashboard over all docs/loen/ runs, or --triage to turn failing runs into proposed next actions (proposals only; never launches loops or edits runs).
Use BEFORE superpowers:brainstorming for any non-trivial work (new module, CLI flag, API change, arch decision). Triggers on "/fix-intent", "/idd", "capture intent", "intent doc". If brainstorming would start on non-trivial work without an approved intent doc in docs/superpowers/intents/, run IDD first. Brainstorm proceeds only when intent Status: approved.