skills.rest
edhoferdian
edhoferdianCOMMUNITY
Explore all skills by this author →
GitHub

code-review-edho-ferdian

Reviews code across five domains with evidence-backed findings and confidence-labeled severity reports.

2|Updated Sep 6, 2026
Search Tags:#static analysis#code review#pull request#security audit#wcag#false positives#github cli
One-click install
npx skills add https://github.com/edhoferdian/EEF --skill code-review-edho-ferdian-edhoferdian
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: code-review-edho-ferdian
Source: https://github.com/edhoferdian/EEF/tree/main/.agents/skills/code-review-edho-ferdian
Command: npx skills add https://github.com/edhoferdian/EEF --skill code-review-edho-ferdian-edhoferdian

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? AI code reviews often produce false positives, inflated severities, and unverified claims that erode maintainer trust. This Skill performs a disciplined senior-engineer review that verifies findings with real tooling, self-corrects through reflection and adversarial critique, and produces a decision-ready report a maintainer can act on without re-checking. ## Core Features & Use Cases - Five-domain review: Covers Code Quality, Security, Performance, Blueprint/Spec Consistency, and Test Quality, plus conditional lenses auto-detected from scope (database, accessibility, RAG, ML engineering, healthcare, agent/LLM stacks). - Ground-truth verification: Runs available linters, type-checkers, test suites, and dependency/secret scanners, labeling every finding with High/Medium/Low confidence instead of guessing. - Reflection + Critique-Correction Loop: A self-audit pass and an adversarial second-reviewer pass (optionally via sub-agents) suppress false positives before the report is shown. - PR Review Mode: Fetches a pull request via the GitHub CLI, treats PR content as untrusted input, and emits an APPROVE / APPROVE-WITH-COMMENTS / REQUEST-CHANGES verdict. - Use Case: Ask it to review a staged diff before merging to main; it detects the stack, runs the five domains plus any activated lenses, verifies findings with your repo's tooling, and writes a saved Markdown report with fixes (full rewrite under 100 lines, surgical patch above). ## Quick Start Ask the assistant to review the current git diff or a specific pull request before merging, for example by saying "review this PR and tell me if it is ready to merge".

Frequently Asked Questions about code-review-edho-ferdian

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review a pull request with AI before merging?▼

Provide a PR number or URL and the Skill fetches the diff, description, and comments via the GitHub CLI, then runs its five-domain review scoped to the change set. It finishes with an APPROVE, APPROVE-WITH-COMMENTS, or REQUEST-CHANGES verdict derived from finding severities.

What does an AI code review check for?▼

It checks five domains: Code Quality, Security, Performance, Blueprint/Spec Consistency, and Test Quality. Conditional lenses add depth for database schemas, accessibility (WCAG 2.2 AA), RAG pipelines, ML training/serving, healthcare/HL7/FHIR code, and agent/LLM tool-calling loops when the scope touches them.

How does the review avoid false positives?▼

Every finding passes a pre-report gate, a six-gate reflection pass, and an adversarial critique-correction loop capped at two rounds. Findings are cross-checked against a false-positive catalogue, and anything unverifiable is dropped, downgraded, or labeled as needing verification.

Does the code review run linters and tests or just read code?▼

It probes the repo for available tooling and runs linters, type-checkers, test suites, and dependency/secret scanners to confirm findings. Tool-confirmed findings get High confidence labels; unverifiable claims are labeled Medium or Low with a note on what would confirm them.

When should I use a dedicated security review instead of a general code review?▼

When the request is entirely about security, such as a security audit of auth, payments, or PHI handling, route to the dedicated security-review skill, which owns the full SEC-01 through SEC-13 criteria. The general review keeps only a slim security summary for quick passes.

Can the review apply fixes automatically?▼

Yes, it uses an adaptive fix mode: files under 100 lines get a full rewrite, larger files get surgical patches. On harnesses like Claude Code it can apply fixes and re-run tests after confirming with the user, and it saves the report as a Markdown file in the repo.

Related Skills

Explore More Software Engineering Skills→
AO-HySAO-HyS

flow-code-review

Review branch diffs against repository standards and specs with ranked actionable findings.

Official
Intermediate
corbat-techcorbat-tech
1

release

Orchestrate a complete software release cycle with versioning, changelog updates, testing, and Git tags.

Official
Advanced
AradotsoAradotso
11

vibe-security-skill

Detect security vulnerabilities in AI-generated code across Supabase, Stripe, and auth flows.

Official
Intermediate
hatch3rhatch3r
26

hatch3r-perf-audit

Measure bundle size, runtime latency, and memory against performance budgets.

Official
Advanced