SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsacceptance

acceptance

Testing & quality

View repository

Updated 10-16 05:01

About this skill

acceptance

Ask questions by target repository .agents/docs/PROJECT.md category; search existing tests / e2e / HTTP / build commands before recommending, generate the global acceptance prompt and necessary scripts, and record whether they can run on this machine. setup does not run this skill on its behalf. It is forbidden to write this skill's Q&A into setup's interview.md. It is forbidden to specify tools without searching. Acceptance files are not a mandatory precheck item.

Pre-checks

Execute according to the "Pre-checks" section of common.md .

Workflow

Target root = current workspace / git root. Confirm first if ambiguous. Subsequent paths are relative to this root.

  1. Read the category of .agents/docs/PROJECT.md ; for full stack, also read the architecture form. If the file is absent: stop, prompt to first setup。
  2. Search first, then recommend. For search items and per-category means, see recommend.md. Do not name tools such as Playwright / Maestro / HTTP scaffolding until the search is complete.
  3. Ask questions using the interview.md tool according to 交互式提问 . At most 5 questions per round; wait until answered before asking the next round. Do not ask what the repository has already answered.
  4. If the interview or repository facts indicate it is a pure prototype or daily work: do not recommend and do not generate executable scaffolding. Still write acceptance-template.md according to .agents/docs/ACCEPTANCE.md(completion criteria empty, note the reason). End.
  5. Layer the search results onto the category means and give a recommendation. Do not invent types that have no convention. Use 交互式提问 to confirm whether to adopt; if the user rejects it, only change within already-searched facts and means that have conventions.
  6. If the recommendation includes a backend HTTP script: execute http-scaffold.md. Choosing this means must collect authentication and has no default. If there are already tests hitting the HTTP entry point, do not add new scaffolding.
  7. Probe whether the commands to be written can run on this machine (visible runner, binary, existing npm/make scripts). Mark those that cannot run as "skip" and write the reason; do not write them as must pass.
  8. Write acceptance-template.md according to .agents/docs/ACCEPTANCE.md. Scripts that need to be landed go only in .agents/docs/acceptance/; for templates see http-scaffold.md and script-templates.md. It is forbidden to write to .agents/scripts/. If ACCEPTANCE.mdalready exists: ask whether to cover or revise; do not silently overwrite. Do not create an empty directory for "no scripts".
  9. Report: written paths, adopted means, commands that can run on this machine, skipped items.

Outputs

Path What to write
.agents/docs/ACCEPTANCE.md Global prompt: commands, whether they can run on this machine, skip markers
.agents/docs/acceptance/ Optional scripts (HTTP / Playwright / Maestro, etc.)

bun can serve as the local runner, but it is forbidden to write bun into the target repository's package.json dependencies (dependencies / optionalDependencies / devDependencies all count).