/g The Contract Codex ← 返回动态 Catalog

Oracle Definition Of Done

Define an oracle — the concrete checks (tests, edge cases, benchmarks, static checks) that decide success — before running the agent task.

难度:intermediate 分类:workflow 来源:where good ideas come from (for coding agents) - Sunil Pai

完整 Prompt(可直接复制)

/goal
GOAL:
Complete Oracle Definition Of Done for a coding-agent workflow repository: Define an oracle — the concrete checks (tests, edge cases, benchmarks, static checks) that decide success — before running the agent task.

CONTEXT:
- Before editing, read the nearest AGENTS.md/CLAUDE.md, current issue or PLAN.md, and any failing logs already in the repo.
- Inspect goal text, progress logs, branch state, and verification artifacts.
- Establish a baseline by running or locating evidence for: `pytest && npm run lint`.

CONSTRAINTS:
- Keep the scope limited to this goal; do not expand into unrelated cleanup.
- Do not weaken tests, delete assertions, or mask errors to make verification pass.
- Respect the repository's AGENTS.md/CLAUDE.md instructions and existing patterns.
- Do not claim a goal is complete without a current audit of the stated contract.
- Pause if the goal text, branch state, or permissions are inconsistent.

DONE WHEN:
- The implementation or documentation directly satisfies: Define an oracle — the concrete checks (tests, edge cases, benchmarks, static checks) that decide success — before running the agent task.
- The verification command or evidence path succeeds: `pytest && npm run lint`.
- The final diff is scoped to the relevant files and has no unrelated formatting churn.

VERIFY:
- Run `pytest && npm run lint` or the closest repo-local equivalent if the exact command is not available.
- Capture before/after evidence for the behavior, metric, report, or artifact involved.
- If verification cannot run locally, stop and report the missing dependency instead of guessing success.

OUTPUT:
- Summarize changed files, key decisions, verification output, and remaining risks.
- Include any follow-up that is required for production rollout or human review.

STOP RULES:
- Pause if secrets, production access, stakeholder decisions, or destructive data operations are required.
- Pause after three failed fix attempts on the same symptom and challenge the root-cause hypothesis.
- Do not mark the goal complete until the current repository state has been audited against DONE WHEN.

来源与证据

原始来源: where good ideas come from (for coding agents) - Sunil Pai

证据摘要: the checks that decide success: tests to add, edge cases, benchmarks, static checks, canary signals.; source: where good ideas come from (for coding agents) - Sunil Pai; type: third-party-tutorial; verification: pytest && npm run lint