메뉴
Use when preparing or running a Claude CLI agentic engineering review through claude -p for repository audits, architecture analysis, system design, security review, large refactors, specs, RFCs, or evidence-grounded long-form engineering analysis.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | claude-agentic-review |
| description | Use when preparing or running a Claude CLI agentic engineering review through claude -p for repository audits, architecture analysis, system design, security review, large refactors, specs, RFCs, or evidence-grounded long-form engineering analysis. |
Use this skill to prepare a Claude CLI review prompt or supervise a Claude CLI agentic pass. Codex main remains final owner of decisions, edits, safety, synthesis, and verification. Treat Claude output as advisory until it is checked against repository evidence and local validation.
Use:
claude -p "{prompt}"
Before sending context to Claude, minimize it to the files and facts needed for the review. Do not send RED or unsanitized sensitive content.
When constructing the Claude prompt, include this operating contract:
Claude Agentic Engineering Skill
You are operating through Claude CLI in fully autonomous agent mode.
Primary objective:
Produce the highest quality result possible while minimizing hallucinations and unnecessary token usage.
Execution protocol:
1. Discovery First
* Never start implementing immediately.
* Build a complete mental model of the repository.
* Identify entrypoints, architecture, dependencies, configuration, build system, deployment flow, test framework, and security boundaries.
2. Evidence-Based Analysis
* Every claim must be grounded in actual files.
* Cite file paths, functions, classes, interfaces, and line ranges when available.
3. Repository Mapping
Before proposing changes create an architecture summary, dependency graph, runtime flow, security model, and build pipeline summary.
4. Planning Phase
Generate findings, risks, opportunities, technical debt, performance issues, and security concerns.
Rank them as Critical, High, Medium, or Low.
5. Implementation Rules
* Prefer minimal changes.
* Preserve backward compatibility.
* Avoid speculative refactors.
* Avoid introducing new dependencies unless justified.
6. Validation
After every implementation run tests, linters, type checks, build validation, and documentation validation when applicable.
7. Output Format
Executive Summary
Findings
Root Causes
Proposed Solution
Implementation Plan
Validation Results
Remaining Risks
8. Token Efficiency
Never dump package-lock, yarn.lock, pnpm-lock, build artifacts, or generated files.
Summarize them instead.
9. Refusal Rule
If information is missing, say exactly what is missing.
Do not guess.
Use this skill for repository audits, architecture and system design, security review, large refactors, specs, RFCs, and detailed evidence-based engineering review.
Use zcode-agentic-builder instead when the goal is fast code generation or
iterative implementation throughput on an already understood codebase.
Improve, audit, simplify, rewrite, or migrate prompts, tool descriptions, agent instructions, and prompt stacks for GPT-5.6 Sol or the GPT-5.6 family. Use for outcome-first prompt design, autonomy boundaries, tool routing, PTC, grounding, verbosity, reasoning effort, and prompt evals.
Apply a deep, design-minded engineering workflow for complex work that needs careful planning, iteration, and simplification.
Use when preparing or running a ZCode GLM-5.2 agentic builder workflow through zcode --prompt for fast implementation, iterative code generation, focused fixes, and validation on an existing repository.
Apply adversarial opposite-analysis to plans, specs, architecture, code changes, and claims. Use when the user asks for adversarial review, opposing analysis, contrarian review, red-team reasoning, or Z.ai and MiniMax cross-checks through the Ralph MCP router.
Review and adjudicate Bugbot, Cursor, Seer, and similar automated PR feedback with local evidence before accepting, fixing, or dismissing findings.
Find real bugs in local code, branch diffs, or named surfaces with evidence-first triage, regression-test guidance, and Codex-native verification.