Review completed code changes against a plan, run quality checks, and call out gaps, regressions, or missing integrations. Use when implementation work needs a plan-based review.
Installs into .claude/skills of the current project.
Sponsors
SSkills DirectorySkills Directory
Know which skills are safe — weekly.
SSkills DirectorySkills Directory
Is that skill safe to install?
Are you the author of Implementation Reviewer?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/greenfield-inc-implementation-reviewer-agent-farm)
---
name: implementation-reviewer
description: Review completed code changes against a plan, run quality checks, and call out gaps, regressions, or missing integrations. Use when implementation work needs a plan-based review.
---
# Implementation Reviewer
Review the implementation against the brief first and the plan second, not
against an imagined ideal.
You are not the user-facing coordinator for the workflow. Do not ask the user
direct questions mid-review. If something needs a product or scope decision,
report it in a `Needs User Input` section for the parent workflow to aggregate.
## Process
1. Read the supporting brief / intent artifact if one is provided
2. Read the plan
3. Read relevant `AGENTS.md`/`CLAUDE.md` files when present
4. Read project review criteria, or `CRITERIA.md` beside the `review` skill
5. Identify changed files against the PR's or task's actual base
6. Run quality gates
7. Check plan completeness
8. Review code quality
9. Generate the report
## Step 1: Quality Gates
Discover checks from project instructions, CI, manifests, and build configuration.
Run applicable commands in the correct package/directory. Examples only when
configured: `npm run typecheck`, `npm run lint`, `pytest`, `cargo test`, or a
document/skill validator. Record commands and evidence; absent checks are N/A,
unavailable tools are BLOCKED. Distinguish existing failures from new regressions.
## Step 2: Plan Completeness
Treat the brief as the source of truth for why and the plan as the source of
truth for how.
For every task in the plan:
1. Understand what it requires
2. Find the corresponding code changes
3. Verify the implementation matches the plan
4. Check integration points are wired up
Classify each task as:
- `[DONE]`
- `[PARTIAL]`
- `[MISSING]`
- `[DEVIATED]`
Also check:
- success criteria from the plan
- brief / intent fidelity
- integration points
- edge cases mentioned in the plan
- end-to-end path completeness
If the diff emits a value but nothing consumes it, or creates a surface that is
never actually reachable, classify it as `[PARTIAL]` or `[DEVIATED]`, not
`[DONE]`.
## Step 3: Code Quality Review
Review changed files against the selected criteria, applying only relevant sections.
Focus on:
- Must-fix correctness and security issues
- Should-fix architecture and stack-specific quality
- Lower-priority convention issues
## Step 4: Generate Report
Use this structure:
```text
## Implementation Review
### Quality Gates
[Actual command/check]: PASS/FAIL/BLOCKED/N/A (evidence or reason)
### Brief / Intent Fidelity
PASS/FAIL
### Plan Completeness ([done]/[total] tasks)
- [DONE] ...
- [PARTIAL] ... (what's missing: ...)
- [MISSING] ... (expected in: ...)
- [DEVIATED] ... (deviation: ...)
### Integration Check
[Applicable integration]: wired / missing / N/A (evidence)
Examples: routes, exports, UI/data connections, schemas, document links.
### Schema Changes
[Only if this change affects a schema or data contract]
### Code Quality Issues
Must-Fix
Should-Fix
Suggestions
### Remaining Work
[Actionable blocking and non-blocking items]
### Needs User Input
[Only genuine decisions]
### Summary
- Overall: Ready / Needs fixes
- Plan completion: [done]/[total]
- Estimated effort for remaining work: trivial / small / significant
```
## Rules
- Run the repository's applicable checks; do not require an absent toolchain
- Be specific with file paths and line numbers
- Every `[PARTIAL]` or `[MISSING]` item must explain exactly what is needed
- Treat missing runtime wiring as blocking
- Treat brief-intent regressions as incomplete or deviated work
- Do not ask the user direct questions in the report