agentleFS
DGouronDGouron/review-flow/.claude/skills/review-followup/SKILL.md

review-followup

Skill44 starsChanged 6 months agoMITView on GitHub

Follow-up review to verify corrections on a MR. Sequential execution to avoid memory spikes. Checks blocking issues, detects new problems, and posts a concise report on GitLab.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read .claude/roles/code-reviewer.md — adopt this profile and follow all its rules.

    line 10
  2. Distinctive:

    Correction verification: Every blocking issue from the previous review must be verified

    line 17
  3. Very specific:

    Check only the critical rules from CLAUDE.md and Clean Code red flags (see /.claude/skills/clean-code/SKILL.md):

    line 138
  4. Distinctive:

    Use the MCP tool to get the discussion threads:

    line 174
  5. Very specific:

    Use the Write tool to persist the full follow-up report locally.

    line 307
  6. Distinctive:

    YOU MUST: use MCP tools to manage threads.

    line 360
  7. Very specific:

    Use the internal keys exactly: security, logic, performance, typeSafety, style, dependencies.

    line 375

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “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

Follow-up Review

Persona

Read .claude/roles/code-reviewer.md — adopt this profile and follow all its rules.

Context

You are: The same demanding reviewer as the initial review, verifying that the requested corrections have been applied.

Your approach:

  • Correction verification: Every blocking issue from the previous review must be verified
  • Regression detection: Did the corrections introduce new problems?
  • Conciseness: Short and actionable report, no pedagogical lessons (already given)
  • Direct: "Fixed" or "Not fixed", no nuances

Activation

This skill activates when the user asks for:

  • "Verify the corrections", "Follow-up review", "/review-followup"
  • "Are the blocking issues fixed?"
  • "Second review", "Re-review"

⚡ Sequential Architecture (Anti Memory-Leak)

CRITICAL: To avoid memory explosion, steps are executed ONE AT A TIME.

┌─────────────────────────────────────────────────────────────────┐
│                    SEQUENTIAL EXECUTION                         │
│                                                                 │
│  [1] Context  →  [2] Verify  →  [3] Scan  →  [4] Threads  →... │
│                                                                 │
│  Each step:                                                     │
│  1. Calls start_agent(jobId, stepName)                          │
│  2. Executes the step                                           │
│  3. Calls complete_agent(jobId, stepName, status)               │
│  4. WAITS before launching the next one                         │
└─────────────────────────────────────────────────────────────────┘

🔧 Available MCP Tools

The MCP server exposes these tools for progress tracking:

ToolUsageArguments
get_workflowRetrieves the workflow state and agent listjobId
start_agentSignals the start of a stepjobId, agentName
complete_agentSignals the end of a stepjobId, agentName, status, error?
set_phaseChanges the workflow phasejobId, phase
get_threadsRetrieves the MR discussion threadsjobId
add_actionAdds an action (resolve thread, reply, comment)jobId, type, ...

The jobId is available via the MCP_JOB_ID environment variable.


⚡ Progress via MCP (MANDATORY)

To enable real-time tracking in the dashboard, use MCP tools at each step:

Phases (only one active at a time):

set_phase(jobId, "initializing")   # At startup
set_phase(jobId, "agents-running") # During verifications
set_phase(jobId, "synthesizing")   # During synthesis
set_phase(jobId, "publishing")     # During GitLab publication
set_phase(jobId, "completed")      # At the end

Steps (one per step):

start_agent(jobId, "context")     # Context retrieval
complete_agent(jobId, "context", "success")
start_agent(jobId, "verify")      # Blocking issues verification
complete_agent(jobId, "verify", "success")
start_agent(jobId, "scan")        # Scan for new problems
complete_agent(jobId, "scan", "success")
start_agent(jobId, "threads")     # GitLab thread management
complete_agent(jobId, "threads", "success")
start_agent(jobId, "report")      # Report generation
complete_agent(jobId, "report", "success")

Workflow

Phase 1: Context Retrieval

Call:

  • set_phase(jobId, "initializing")
  • start_agent(jobId, "context")
  1. Identify the MR from the provided URL or number
  2. Use get_threads(jobId) to retrieve discussion threads
  3. Read previous comments to identify:
    • Blocking issues (🚨)
    • Important issues (⚠️)
  4. Retrieve the current diff to see the modifications

Call: complete_agent(jobId, "context", "success")


Phase 2: Blocking Issues Verification

Call:

  • set_phase(jobId, "agents-running")
  • start_agent(jobId, "verify")

For EACH blocking issue identified:

StatusCriterion
✅ FIXEDThe code has been modified according to the requested correction
⚠️ PARTIALLYFixed but with reservations or a different approach
❌ NOT FIXEDThe problem persists in the current code

Call: complete_agent(jobId, "verify", "success")


Phase 3: Quick Scan for New Problems

Call: start_agent(jobId, "scan")

Check only the critical rules from CLAUDE.md and Clean Code red flags (see /.claude/skills/clean-code/SKILL.md):

RuleWhat to check
Type anyNo new any introduced
Type assertionsNo new as Type or as any
Law of DemeterNo chaining a.b.c.d
Imports@/ aliases used
TestsNew business logic = new test
Function size (Clean Code)No new function > 20 lines or with 4+ arguments
Flag arguments (Clean Code)No new boolean parameter switching behavior
Naming (Clean Code)No new abbreviation (ctx, idx, gw), intention-revealing identifiers
Comments (Clean Code)No new comment rephrasing the code, no commented-out code
Magic numbers (Clean Code)New literals (> 7, * 86400) extracted to named constants

Do not check (out of scope for follow-up):

  • Global architecture
  • Strategic DDD
  • Framework-specific performance
  • Full SOLID
  • Full Clean Code audit (only red flags above; for a complete pass, see review-front)

Call: complete_agent(jobId, "scan", "success")


Phase 4: Thread Management via MCP

Call: start_agent(jobId, "threads")

⚠️ CRITICAL - USE MCP TOOLS

Threads are managed via MCP tools, no longer via JSON file.

Step 1: Retrieve threads

Use the MCP tool to get the discussion threads:

get_threads(jobId)

Returns a list of threads with:

  • id: Thread identifier
  • file: Related file
  • line: Comment line
  • status: "open" or "resolved"
  • body: Comment content

Step 2: Add actions for each fixed thread

For each FIXED issue, add actions via MCP:

Resolve a thread:

add_action(jobId, "THREAD_RESOLVE", threadId="xxx", message="✅ Fixed")

Reply to a thread:

add_action(jobId, "THREAD_REPLY", threadId="xxx", message="✅ Fixed - Short description")

Post a general comment:

add_action({ jobId: JOB_ID, type: "POST_COMMENT", body: "Report content" })

Post an inline comment (new problem):

add_action({ jobId: JOB_ID, type: "POST_INLINE_COMMENT", filePath: "path/file.ts", line: 42, body: "..." })

Usage Rules

ActionWhen to use
THREAD_RESOLVEBlocking issue fixed → Resolve the thread
THREAD_REPLYNeed to clarify the fix before resolving
POST_COMMENTPost the global follow-up report
POST_INLINE_COMMENTNew blocking problem detected → Inline comment on the diff

IMPORTANT:

  • ✅ Use MCP tools get_threads and add_action
  • ✅ Actions are queued and executed after the review
  • ❌ Do NOT use JSON files (deprecated legacy system)
  • ❌ Do NOT call glab api or gh api directly

Call: complete_agent(jobId, "threads", "success")


Phase 5: Report Generation

Call:

  • set_phase(jobId, "synthesizing")
  • start_agent(jobId, "report")

Follow-up report structure:

# Follow-up Review - MR [Number]

**Date**: [YYYY-MM-DD]
**Previous review**: [Date of the first review]

---

## Blocking Issues Verification

| # | Issue | Status | Comment |
|---|-------|--------|---------|
| 1 | [Short description] | ✅/⚠️/❌ | [Note if needed] |
| 2 | ... | ... | ... |

---

## New Problems Detected

### [If problems found]

🚨 **[Title]**
📍 `file.ts:line`
[Short description + correction]

### [If no problems]

No new problems detected in the modifications.

---

## Final Verdict

| Criterion | Status |
|-----------|--------|
| Blocking issues fixed | X/Y |
| New blockers | X |
| **Ready to merge** | ✅ Yes / ❌ No |

### Required Actions (if not ready)

1. [Action 1]
2. [Action 2]

Call: complete_agent(jobId, "report", "success")


Publication

Call: set_phase(jobId, "publishing")

Order of Operations (STRICT)

┌─────────────────────────────────────────────────────────────────┐
│  1. Retrieve threads via get_threads(jobId)                     │
│  2. Verify each issue (fixed/not fixed)                         │
│  3. Add actions via add_action(jobId, ...)                      │
│  4. Post the global report on the MR                            │
│  5. Post inline comments for NEW problems only                  │
└─────────────────────────────────────────────────────────────────┘
  1. Manage existing threads via MCP tools (see Phase 4 section)
  2. Save the MD in /.claude/reviews/[YYYY-MM-DD]-MR-[ID]-followup.md

    Use the Write tool to persist the full follow-up report locally. The orchestrator reads this file to confirm the followup ran successfully — if it is missing, the job is reported as failed.

  3. Post the report on the MR:
    add_action({ jobId: JOB_ID, type: "POST_COMMENT", body: "<report content>" })
  4. Inline comments for new problems only (via POST_INLINE_COMMENT)

Inline Comments for New Problems

Post inline comments via MCP ONLY if:

  • A NEW blocking problem is detected → POST_INLINE_COMMENT
add_action({
  jobId: JOB_ID,
  type: "POST_INLINE_COMMENT",
  filePath: "path/to/file.ts",
  line: 42,
  body: "🚨 **[NEW] Problem title**\n\n📍 `file.ts:42`\n\n[Short description + correction]"
})

Exit Commands

Call: set_phase(jobId, "completed")

At the end of the followup:

🔄 Follow-up Review - MR [ID]

📊 Blocking issues: X/Y fixed
✅ Threads resolved: X (via add_action MCP)
🆕 New problems: X

[REVIEW_STATS:blocking=X:warnings=X:suggestions=X:score=X:categories=security=N,logic=N,performance=N,typeSafety=N,style=N,dependencies=N]

✅ READY TO MERGE
or
❌ CORRECTIONS REQUIRED:
   - [Action 1]
   - [Action 2]

⚠️ IMPORTANT:

  • The [REVIEW_STATS:...] line is MANDATORY for automated tracking.
  • The "Threads resolved" count corresponds to the number of add_action(jobId, "THREAD_RESOLVE", ...) calls.
  • YOU MUST use MCP tools to manage threads.

Category breakdown (categories= segment)

Classify every finding still open or newly raised (blocking + important + suggestion) into exactly one of the six dashboard categories, then emit the count per category in the marker. The six counts must sum to blocking + warnings + suggestions. Map by the nature of the finding:

Finding natureCategory
Security / credential / auth issuesecurity
Business rule, architecture or SOLID violationlogic
Performance / inefficiencyperformance
Type safety (any, unsafe assertion, weak typing)typeSafety
Readability, naming, testing, code qualitystyle
Outdated / vulnerable / misused dependencydependencies

Use the internal keys exactly: security, logic, performance, typeSafety, style, dependencies. Set any unused key to 0. Unknown keys are ignored by the tracker.


Limited Read/Write Mode

This skill can:

  • ✅ Read, analyze, compare code
  • ✅ Post reports and comments on GitLab
  • ✅ Use MCP tools for tracking and thread management
  • ❌ Modify the project source code

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.