agentleFS
solo-istsolo-ist/prose/.agents/skills/review-feedback/SKILL.md

review-feedback

Skill8 starsChanged 5 months agoMITView on GitHub

Pragmatic analysis of PR feedback. Validates issues, assesses severity, and recommends actions (fix, defer, or dismiss).

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the GitHub MCP to get PR information and all review comments:

    line 29
  2. Distinctive:

    Is it real?: - Read the code, understand the context, reproduce if possible

    line 64
  3. Distinctive:

    If the user approves, create GitHub issues for deferred items:

    line 122
  4. Distinctive:

    Validate before acting: - Don't assume feedback is correct.

    line 130
  5. Distinctive:

    Document deferrals: - Create tickets so nothing is lost.

    line 134
  6. Distinctive:

    All functional issues addressed

    line 140

“Common” is measured: kept by at least 6 in 10 of the 4 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 PR Feedback

Pragmatic analysis of PR review comments. Validates concerns, assesses severity, and provides actionable recommendations.

Philosophy

Pragmatism over compliance. This skill does not blindly accept all feedback:

  • Validate that issues are real and reproducible
  • Weigh effort vs. impact
  • Recommend deferring low-priority issues to follow-up tickets
  • Focus on shipping quality code, not perfect code

Usage

/review-feedback <pr-number>

Workflow

1. Fetch PR and Review Comments

Use the GitHub MCP to get PR information and all review comments:

mcp__github__pull_request_read (method: "get", owner: "solo-ist", repo: "prose", pullNumber: <pr-number>)

Then fetch review comments (inline code comments):

mcp__github__pull_request_read (method: "list_reviews", owner: "solo-ist", repo: "prose", pullNumber: <pr-number>)
mcp__github__pull_request_read (method: "get_review_comments", owner: "solo-ist", repo: "prose", pullNumber: <pr-number>)

And general PR comments:

mcp__github__pull_request_read (method: "get_comments", owner: "solo-ist", repo: "prose", pullNumber: <pr-number>)

2. Categorize Feedback

Group each piece of feedback into categories:

CategoryDescriptionExamples
BlockingSecurity, crashes, data lossSQL injection, null pointer, file corruption
FunctionalBugs, broken featuresFeature doesn't work, edge case failure
Code QualityStyle, patterns, maintainabilityNaming conventions, design patterns
NitpicksPreferences, minor suggestionsFormatting, comment wording
QuestionsClarifications, not actionable"Why did you choose X?"

3. Validate Each Issue

For each piece of feedback, investigate:

  1. Is it real? - Read the code, understand the context, reproduce if possible
  2. Is it in scope? - Does it relate to this PR's changes, or is it pre-existing?
  3. Is it accurate? - Is the reviewer's assessment correct?

Read the actual code files to validate. Don't trust summaries.

4. Assess Severity & Impact

For validated issues:

DimensionLevels
SeverityCritical / High / Medium / Low
EffortQuick fix (< 5 min) / Moderate (< 30 min) / Significant refactor
RiskWhat happens if we don't fix it?

5. Generate Recommendations

Output this structured analysis:

## PR Feedback Analysis: #<number>

### Summary
<1-2 sentence overview of feedback and recommendation>

### Recommendation: [MERGE | FIX REQUIRED | NEEDS DISCUSSION]

---

### Must Fix Before Merge
| Issue | Severity | Effort | Rationale |
|-------|----------|--------|-----------|
| ... | Critical | Quick | ... |

### Consider Fixing
| Issue | Severity | Effort | Recommendation |
|-------|----------|--------|----------------|
| ... | Medium | Moderate | Fix now / Defer |

### Defer to Follow-up
| Issue | Rationale | Suggested Ticket |
|-------|-----------|------------------|
| ... | Out of scope | "Refactor X for consistency" |

### Dismissed
| Feedback | Reason |
|----------|--------|
| ... | Style preference, not a bug |

---

### Follow-up Tickets (if applicable)
Ready-to-create issue descriptions for deferred items.

6. Optional: Create Follow-up Issues

If the user approves, create GitHub issues for deferred items:

mcp__github__issue_write (method: "create", owner: "solo-ist", repo: "prose", title: "...", body: "...")

Key Principles

  1. Validate before acting - Don't assume feedback is correct. Read the code.
  2. Scope matters - Reject scope creep politely. Pre-existing issues belong in separate tickets.
  3. Effort vs Impact - A 2-hour fix for a cosmetic issue isn't worth blocking a merge.
  4. Ship it - Perfect is the enemy of good. Working code today beats perfect code never.
  5. Document deferrals - Create tickets so nothing is lost. Defer, don't ignore.

Analysis Guidelines

When to recommend MERGE:

  • No blocking issues
  • All functional issues addressed
  • Remaining feedback is cosmetic or out-of-scope

When to recommend FIX REQUIRED:

  • Blocking issues exist (security, crashes, data loss)
  • Functional bugs that affect users
  • Tests failing or missing for critical paths

When to recommend NEEDS DISCUSSION:

  • Significant architectural concerns
  • Scope disagreement between reviewer and author
  • Trade-offs that need team input

Example Output

## PR Feedback Analysis: #42

### Summary
Reviewer flagged 5 items: 1 valid bug, 2 style preferences, 1 pre-existing issue, 1 question.
The bug is a quick fix; other items can be deferred or dismissed.

### Recommendation: FIX REQUIRED

---

### Must Fix Before Merge
| Issue | Severity | Effort | Rationale |
|-------|----------|--------|-----------|
| Null check missing in `handleSave` | High | Quick | Crashes on new files |

### Consider Fixing
| Issue | Severity | Effort | Recommendation |
|-------|----------|--------|----------------|
| Variable naming (`tmp` vs `tempFile`) | Low | Quick | Fix now - easy win |

### Defer to Follow-up
| Issue | Rationale | Suggested Ticket |
|-------|-----------|------------------|
| Refactor file service to use Result type | Pre-existing code, out of scope | "Refactor file service error handling" |

### Dismissed
| Feedback | Reason |
|----------|--------|
| "Consider using lodash here" | Adds 70kb dep for one function, native works fine |
| "Why not use React Query?" | Question, not actionable. Current approach works. |

---

### Follow-up Tickets

**Title:** Refactor file service error handling
**Body:**
The file service currently uses try/catch with mixed return types. Consider:
- Using a Result<T, E> pattern for explicit error handling
- Consolidating error types

Related: PR #42 review feedback

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.