agentsclimarketplace

Intent rich pr

Skill lavrovpy/al-prompts/skills/intent-rich-pr

Create GitHub pull requests with intent-rich descriptions that explain why the change exists, link ticket or issue context, summarize the diff, capture reviewer-facing decision rationale, identify review risk, and include validation. Use when the user asks to create, open, draft, or write a PR/MR and wants reviewers or AI review agents to understand the underlying motivation, reasoning, tradeoffs, risk, or issue context.From its SKILL.md

Install
npx -y skills add lavrovpy/al-prompts --skill intent-rich-pr

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.3 KB, as published. Nobody here has run it

Intent-Rich PR

Overview

Use this skill to create a pull request whose description gives reviewers the missing context behind the diff. Treat the PR body as review intake: it should make intent, evidence, and risk visible before a human or AI reviewer inspects changed files.

Workflow

  1. Confirm the repository state.

    • Run git status -sb.
    • Identify the current branch, upstream, base branch, commits, and changed files.
    • If the worktree has unrelated uncommitted changes, stop and ask what belongs in the PR.
  2. Collect motivation before drafting.

    • Ask for or derive the original user request, ticket, GitHub issue, Jira issue, Linear issue, design doc, or incident context.
    • If a ticket or issue exists, include its link in the PR body.
    • If a GitHub issue is referenced and accessible, fetch its title/body with gh issue view or the available GitHub connector.
    • If an external ticket is referenced but unavailable, ask the user to paste the relevant description or summarize it.
    • Do not invent ticket context. If there is no ticket, say so in the PR body and use the user request as the motivation.
  3. Inspect the implementation.

    • Review git log <base>..HEAD, git diff --stat <base>...HEAD, and the relevant diff.
    • Identify what changed, why the chosen implementation fits the motivation, and what reviewer risks remain.
    • Prefer concrete file/module names over vague summaries.
  4. Assess review risk.

    • Match review depth to blast radius, not authorship. Call out whether the change touches low-risk boilerplate, user-facing behavior, data handling, auth, payments, security, migrations, CI, core business logic, or other load-bearing paths.
    • Flag high-maintenance review signals: large diffs, many touched files, broad generated output, many test edits, CI/lint/coverage changes, or unclear ownership.
    • Read test and CI changes carefully. Note when tests were rewritten, skipped, weakened, or updated to match new behavior.
    • Treat AI review output as a useful signal, not a merge verdict. A human owns the merge decision for any meaningful blast radius.
  5. Capture decision rationale.

    • Include reviewer-facing reasoning only: important decisions, rejected alternatives, tradeoffs, assumptions, and course corrections that explain why the diff looks this way.
    • Do not include raw hidden chain-of-thought or speculative inner monologue.
    • For agent-authored work, preserve the useful plan, intent, and alternatives so the reviewer does not have to reconstruct why the change exists from the diff alone.
    • If the implementation changed direction during the work, summarize the reason and final choice.
    • If nothing unusual happened, keep this section short.
  6. Draft the PR body.

    • Read references/pr-body-guide.md before drafting.
    • Use real Markdown prose with clear headings.
    • Put reviewer context and motivation before the diff summary.
    • Include review focus when the change has notable blast radius, tricky tests, security implications, CI changes, or any area where reviewers should spend extra attention.
    • Explicitly tell AI/code-review agents to read the description before reviewing when that is relevant.
    • Include validation commands and results, or state what was not run.
  7. Create the PR.

    • Default to a draft PR unless the user explicitly asks for a ready PR.
    • Prefer gh pr create --body-file <temp-file> or the available GitHub connector so Markdown renders correctly.
    • Use a concise title that reflects the whole branch.
    • After creation, return the PR URL, base branch, head branch, draft status, and any missing validation.

PR Body Requirements

Every intent-rich PR body should include:

  • Reviewer context: A short instruction telling reviewers what context matters before reading the diff.
  • Motivation: The user request, ticket/issue summary, and ticket/issue link when available.
  • Decision rationale: Key implementation reasoning that is useful for review.
  • Review focus: The blast radius, risky files, test/CI concerns, or why the change is low-risk.
  • What changed: A concise diff-oriented summary.
  • Validation: Commands/checks run and their results.

Optional sections:

  • Risk or follow-up: Known gaps, intentionally deferred work, migration concerns, or manual checks.
  • Screenshots or artifacts: Only when relevant to UI or generated asset work.

Quality Rules

  • Keep the PR description useful, not ceremonial.
  • Do not pad the body with generic boilerplate.
  • Do not claim that a ticket, issue, test, or review happened unless verified.
  • Preserve the user's stated intent even when the diff summary is shorter.
  • Make the causal chain clear: ticket/user need -> implementation choice -> changed files -> validation.
  • Make the review allocation clear: cheap deterministic checks, AI review signals, and human attention should each have an obvious role when relevant.

Keep looking

Skills are one crate of 626,045. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.