Turn a vague task into a testable definition of done and generate an executable goal-test script for it, optionally with a bounded retry loop around a headless agent. Use when the user wants to run an agent in a loop, asks how to know when an agent task is finished, or says a goal like "improve X" needs to become checkable. Do NOT use for building eval suites over many cases — that is evals-bootstrap.
インストール方法を見るインストールする前に、エージェントに与えられる指示の中身を確認できます。
Theory: The four parts of a loop and the goal test build page. "Improve the error handling" cannot terminate a loop, because nothing can ever say it is finished. The goal must be phrased so that a program — not a person, not the model — returns true or false against it.
make test is
better than one that reimplements it.goal-test.sh (or .py if the checks are easier there) in the
project root: runs every check, prints one line per check with pass/fail,
exits 0 only when all pass. Keep it under ~40 lines; it must run in
seconds and be safe to run repeatedly.#!/usr/bin/env bash
# goal-loop.sh — run a headless agent until the goal test passes or attempts run out.
set -u
MAX_ATTEMPTS=5
for i in $(seq 1 "$MAX_ATTEMPTS"); do
FEEDBACK=$(./goal-test.sh 2>&1) && { echo "done in $i attempt(s)"; exit 0; }
claude -p "Goal: <the goal>. The goal test currently fails with:
$FEEDBACK
Fix the code so the goal test passes." \
--permission-mode acceptEdits --output-format json > ".attempt-$i.json"
done
echo "goal test still failing after $MAX_ATTEMPTS attempts — falling back to a human"
exit 1
Adjust the agent command to whatever CLI the user runs. Keep the three
brakes visible and named: the checker outside the model (goal-test.sh),
the stop rule (test passes), the budget (MAX_ATTEMPTS, plus a dollar cap
read from the JSON output if they want one).
まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Audit an agent's context layout against the four places: system prompt, tools, history, tail. Use when the user asks to audit, review or fix their agent's context, prompt caching, token spend per turn, or CLAUDE.md / AGENTS.md layout. Read-only: reports findings and a fix list, changes nothing. Do NOT use for writing evals (evals-bootstrap) or for loop design (goal-test).
日本語の概要は準備中です。原文の説明を表示しています。
Scaffold a first eval suite for an agent: mine real failures into cases, write behavioural checks over traces, and generate the runner. Use when the user wants evals, regression tests for an agent, a golden set, or asks how to know a prompt/model change did not break things. Do NOT use for a single task's done-check (goal-test) or for auditing context (context-audit).
日本語の概要は準備中です。原文の説明を表示しています。
Find the decisions in a pipeline that do not need the expensive model and propose the gate for each: a rule, a classic classifier, or a small model, with fail-closed routing. Use when the user asks to cut model costs or latency, says the big model handles everything, or wants a triage / routing / filter layer in front of an agent. Analysis plus an optional baseline scaffold. Do NOT use for auditing context layout (context-audit) or for building eval suites (evals-bootstrap).
日本語の概要は準備中です。原文の説明を表示しています。
Audit an agent's harness against the four rings: containment, guides, sensors, permissions. Use when the user asks whether their agent setup is safe, wants a review of sandboxing / permissions / hooks / AGENTS.md guardrails, or asks what a bad session could break. Read-only: reports the blast radius and a hardening list, changes nothing. Do NOT use for context layout (context-audit) or cost routing (gate-check).
日本語の概要は準備中です。原文の説明を表示しています。
Import a large archive into the vault in controlled batches: triage what is worth ingesting, process oldest first, checkpoint after every batch, and keep cost visible. Use this skill whenever the user wants to import years of bookmarks, an export from another notes app, a downloads folder, a chat history dump, or says "backfill", "bulk import", "process my archive". Do NOT use for ingesting one or a few new sources, which is second-brain-ingest.
日本語の概要は準備中です。原文の説明を表示しています。
Trace where the user's recorded position on something has shifted: superseded claims, contradictions between sources of different dates, and what evidence moved them. Use this skill when the user asks what they have changed their mind about, how their thinking on a topic developed, what they used to believe, or wants a quarterly look back at their own positions. Do NOT use for finding contradictions between sources with no time dimension, or for a current-state answer.
日本語の概要は準備中です。原文の説明を表示しています。