CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

check-pr

Waits for a PR's checks and reviews to finish, then reports the unresolved review threads, the failing checks and the gaps in the PR description. Use when the user asks to check a PR, whether a PR is ready, or what the Greptile review found.

Quality

80%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./.agents/skills/check-pr/SKILL.md
SKILL.md
Quality
Evals
Security

Check PR

Reports what is left on a pull request in storybookjs/storybook. It changes nothing. To act on the review threads, use handle-pr-comments.

queries.md has the commands and GraphQL queries for the steps below.

Steps

  1. Find the PR. Use the number the user gave, else the PR of the current branch: gh pr view --json number -q .number.

  2. Wait for the checks. Poll statusCheckRollup every 30 seconds until no check is pending or in progress. Wait for "Greptile Review" too, when Greptile reviews the head commit. CI runs only when the PR has a ci:* label (see Pull requests), so a PR without one has few checks.

  3. Collect. For each area, list what needs action:

    • Checks: each failing check and the reason. Danger checks the title, the labels and the "Manual testing" section; see Pull requests.
    • Review threads: each unresolved thread, with its author, file and line. Use GraphQL, because only GraphQL gives a thread's id and whether it is resolved.
    • PR comments: each comment on the PR itself (comments in gh pr view) that asks for something and has no answer yet, from humans and bots.
    • Greptile summary: Greptile edits one summary comment on each review. Read its latest version (newest updated_at), with the confidence score and the findings.
    • Description: placeholders, an empty "Manual testing" section, and "How to review" links to commits that are no longer on the branch.
  4. Report. One table with these columns: area, item, author, and status. The status is one of: needs action, for information, or already fixed by a later commit. Put human comments before bot comments. End with the Greptile confidence score and the number of items that need action.

Done when all checks are finished and each unresolved thread is in the table.

Repository

storybookjs/storybook

Path

.agents/skills/check-pr/SKILL.md

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.