agentleFS
MicrosoftKnown publishermicrosoft/powerplatform-build-tools/.claude/skills/review/SKILL.md

review

Skill314 starsChanged 7 months agoMITView on GitHub

Autonomous PR review — reads diff, cross-references knowledge base, posts inline comments, and leaves an overall verdict.

192 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use close-spec, design-with-images and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • dzhngSkill

    close-spec

    When a spec under specs/<feature>/ is done shipping (a write-spec build, or any planned task that produced a spec), archive it to specs/done/ and rewrite it from a build-plan into a durable rationale — the why, the principles, the invariants — pointing back to the real code for the how. Use when implementation has landed and the plan no longer matches what shipped, or the user says a feature/spec is finished. Pairs with [write-spec](../write-spec/SKILL.md) (the plan this closes) and [review](../review/SKILL.md) (run before closing).

    Also in 5 projects that use this

  • dzhngSkill

    design-with-images

    Explore UI and visual design with image generation, then iterate the real implementation against the user-selected concept. Use when brainstorming visual directions, offering mockup variants, or implementing an approved generated design; do not stop at the mockups or the first code approximation.

    Also in 5 projects that use this

  • LambenthanSkill

    daily-arxiv

    每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新

    Also in 5 projects that use this

  • LambenthanSkill

    exp-eval

    实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

    Also in 5 projects that use this

  • +3 more from dzhng
  • +5 more from Lambenthan

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 157 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    If last sync was > 7 days ago or file missing: run /knowledge-sync inline, then continue.

    line 39
  2. Distinctive:

    Check the diff for:

    line 96
  3. Very specific:

    No flat minimatch@^3.x override — causes infinite npm resolution loop

    line 98
  4. Very specific:

    No new direct dependency that should be in bundleDependencies

    line 102
  5. Distinctive:

    No corresponding merged main PR exists with the same changes

    line 110
  6. Distinctive:

    Verify the paired main PR exists:

    line 112
  7. Distinctive:

    Run in parallel to inform the review:

    line 123
  8. Distinctive:

    No secrets, tokens, or credentials committed

    line 163

“Common” is measured: kept by at least 6 in 10 of the 157 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Review Agent — Autonomous PR Review

Reads a PR diff, cross-references the full knowledge base, posts inline comments on specific issues, and leaves an overall verdict. Zero user input required.

Invoke as:

  • /review <pr-number-or-url> — review a specific PR
  • /review — review the open PR on the current branch

For creating PRs use /create-pr. For CLI version bumps use /pac-cli-update.


Core rule: never ask, always decide

ConditionAction
Line-level issue (wrong pattern, bug risk, breaking change)Post inline comment on that line
File-level concern (missing test, architecture violation)Post inline comment on first line of file
Overall concern (security, missing checklist item)Add to overall review body
No issues in a sectionSkip the section — don't pad the review

Step 1 — Sync knowledge if stale

grep "## Last sync" memory/ado-knowledge.md 2>/dev/null | tail -1

If last sync was > 7 days ago or file missing: run /knowledge-sync inline, then continue.


Step 2 — Resolve PR and classify type

# Resolve PR number, head SHA, base branch, changed files
gh pr view <number-or-url> --json number,title,body,headRefOid,baseRefName,author,url 2>&1
gh pr checks <number-or-url> 2>&1
gh pr view <number-or-url> --json baseRefName,files,author \
  --jq '{base:.baseRefName, author:.author.login, files:[.files[].path]}' 2>&1

Capture: PR_NUMBER, HEAD_SHA, BASE_BRANCH.

Classify the PR type — determines which checks run:

Changed filesPR typeChecks
nuget.json + extension/overview.md onlyPAC CLI bumpStep 3a
package.json + package-lock.json onlynpm dep updateStep 3b
package.json + ALL task.json filesNode target updateStep 3b + Step 5
src/tasks/*/index.ts or selective task.jsonFeature / new inputStep 5 (full)
Base = release/stable (any type)Release branch PRStep 3c first, then type check

Step 3 — Type-specific checks (run the section matching the PR type)

3a — PAC CLI bump checks (nuget.json changed)

# Verify both packages have the same version
node -e "
const n = require('./nuget.json');
const vers = n.packages.map(p => p.version);
console.log('versions:', [...new Set(vers)].join(', '));
console.log(vers.every(v => v === vers[0]) ? 'OK: match' : 'MISMATCH');
" 2>&1

# Verify the version exists on nuget.org
curl -s "https://api.nuget.org/v3-flatcontainer/microsoft.powerapps.cli/index.json" \
  | node -e "const d=JSON.parse(require('fs').readFileSync('/dev/stdin','utf8')); \
    const v=require('./nuget.json').packages[0].version; \
    console.log(d.versions.includes(v)?'nuget: found':'nuget: NOT FOUND')" 2>&1

Request changes immediately if:

  • nuget.json packages have mismatched versions
  • Version does not exist on nuget.org
  • overview.md has no entry for the new version
  • {{NextReleaseVersion}} placeholder was removed from overview.md

3b — npm dependency checks (package.json changed)

Check the diff for:

  • No flat minimatch@^3.x override — causes infinite npm resolution loop
  • ajv override stays at ^6.x — v8 breaks ESLint
  • New overrides are at minimum required version (not over-pinned)
  • inBundle: true entries in package-lock.json changes are intentional (overrides can't reach them)
  • No new direct dependency that should be in bundleDependencies

3c — Release/stable branch checks (base == release/stable)

Release PRs must be version bumps only. Request changes immediately if:

  • Contains .ts logic changes (not just version string changes)
  • Adds new dependencies not present in the corresponding main PR
  • No corresponding merged main PR exists with the same changes

Verify the paired main PR exists:

gh search prs --repo microsoft/powerplatform-build-tools \
  --state merged --base main "<version-or-keyword>" 2>&1

Step 4 — Research context

Run in parallel to inform the review:

# Find similar past PRs (2-3 keywords from PR title)
gh search prs --repo microsoft/powerplatform-build-tools \
  --state merged --limit 8 "<keyword1> <keyword2>" 2>&1

# For top matches, read the body:
gh pr view <past-pr-number> --json title,body,mergedAt 2>&1

Also check memory/ado-knowledge.md for related ADO work items. Skip ADO query for pure PAC CLI or dependency PRs — rarely adds value.


Step 5 — File-level review (feature/bug fix PRs, or any PR touching src/)

Fetch the full diff:

gh pr diff <PR_NUMBER> 2>&1

For each changed file apply:

File patternWhat to check
src/tasks/*/index.tstl.getInput() used correctly; errors reach tl.setResult(TaskResult.Failed, ...); no pac args assembled here
src/tasks/*/task.jsonInput names unchanged (renaming breaks customer pipelines); type correct; required set correctly
src/host/BuildToolsHost.tsALL 32 tasks affected — flag blast radius
src/params/auth/getCredentials.tsAll four auth types still work: UsernamePassword, SPN, ManagedIdentity, WorkloadIdentity
src/params/auth/getEnvironmentUrl.ts4-level fallback preserved: task input → pipeline variable → connection → default
extension/task-metadata.jsonLIVE GUIDs must never change
extension/extension-manifest.jsonPublisher and extension ID must not change
gulp/pack.mjstar imported via createRequire, not ESM import
test/unit-test/**New behaviour is tested; mocks use rewiremock

Security (every PR):

  • No secrets, tokens, or credentials committed
  • No --no-verify or commit hook bypasses

Architecture alignment:

  • Pipeline logic stays in build-tools; pac CLI arg construction stays in cli-wrapper
  • New tasks follow IIFE → isRunningOnAgent() → main() → cli-wrapper pattern

Known failure patterns from knowledge base:

  • WhoAmI locale failure (ADO #4846644): flag any change near whoAmI or locale handling
  • AAD/OAuth authority (ADO #4863652): resolveCloudInstance() map must be preserved
  • PVA import failure (IcM 604312672): pac CLI version bumps should reference changelog

Step 6 — Post inline comments

HEAD_SHA=$(gh pr view <PR_NUMBER> --json headRefOid --jq .headRefOid)

gh api repos/microsoft/powerplatform-build-tools/pulls/<PR_NUMBER>/comments \
  --method POST \
  --field body="<comment>" \
  --field commit_id="$HEAD_SHA" \
  --field path="<file-path>" \
  --field line=<line-number> \
  --field side="RIGHT" 2>&1

Format: ⚠️ <risk> / 🚨 Breaking: <impact> / 💡 Suggestion: / ❓ <question>

Max 15 inline comments — consolidate minor points into the overall review body.


Step 7 — Post overall review and verdict

gh pr review <PR_NUMBER> --comment --body "$(cat <<'EOF'
## Review

**Reviewed by:** Claude Code (autonomous review agent)
**PR type:** <PAC CLI bump / npm dep update / feature / release branch>
**Similar PRs:** <titles + numbers, or "none">
**ADO context:** <relevant items, or "none">

### Breaking changes
<task.json, GUID, or interface changes — or "None detected">

### Security
<!-- Omit if no findings -->

### Architecture
<!-- Omit if no findings -->

### Dependencies
<!-- Omit if package files unchanged -->

### Suggestions
<!-- Omit if none -->

### Verdict
**✅ Approve** / **🔄 Request changes** / **💬 Comment** — <one line reason>
EOF
)" 2>&1

# Then execute the verdict:
gh pr review <PR_NUMBER> --approve 2>&1
# or
gh pr review <PR_NUMBER> --request-changes --body "<blocker>" 2>&1

Verdict decision table

ConditionVerdict
LIVE GUID changedRequest changes — critical breaking
task.json input renamedRequest changes — breaks customer pipelines
Secret/credential in diffRequest changes — security
CI failing on non-functional-test stepRequest changes
release/stable PR has logic changesRequest changes — version bumps only on release branch
release/stable PR missing paired main PRRequest changes
nuget.json package version mismatchRequest changes — CLI update incomplete
PAC CLI version not on nuget.orgRequest changes — version doesn't exist
Only non-blocking suggestionsApprove with inline suggestions
PAC CLI bump verified cleanApprove
No findingsApprove
Intent unclearComment

Never:

  • Approve a PR with a LIVE GUID change, renamed input, or committed secret
  • Block a PR because functional tests fail locally (expected — require live credentials)
  • Re-flag elliptic LOW vuln (known accepted risk, no patch exists)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.