Skill Claude CodeCodex
Measure and investigate repository-level structure, complexity, duplication, dependencies, and composition using deterministic evidence. Use when the user asks for codebase health, maintainability, architecture or complexity analysis, structural hotspots, technical-debt evidence, or the codebase-level analogue of…
Skill Claude CodeCodex
Perform a professional, comprehensive code review of completed work. Use when the user asks to perform a code review. Focus on line-specific findings, risk-based depth, and explicit attention to correctness, safety, tests, and scientific C/C++/Python concerns.
Skill Claude CodeCodex
Simplifies and refines working code for clarity, consistency, and maintainability while preserving exact functionality. Use only when the user explicitly asks for simplification, refactoring, leaner code, or a holistic cleanup pass.
Skill Claude CodeCodex
Use this skill whenever creating or staging a git commit.
Skill Claude CodeCodex
Audit an implementation diff against a frozen contract before normal code review. Use when the user explicitly asks for drift audit, authorization audit, scope audit, or to check whether a completed implementation stayed inside an implementation-plan or scoped-implementation slice.
Skill Claude CodeCodex
Create a compact, high-signal handoff for continuing in another chat or coding agent. Use when the user says "handoff".
Skill Claude CodeCodex
Create a narrow, auditable implementation plan with frozen acceptance slices. Use only when the user explicitly asks for this skill, an implementation plan, or a plan-first workflow before coding.
Skill Claude CodeCodex
Run deterministic linters over a change and report only the findings it introduced. Use when the user asks to lint, run static analysis, or check formatting, and as the mechanical hygiene step before drift-audit and code-review in the plan → implement → audit → review → commit chain. Reports a missing linter as…
Skill Claude CodeCodex
Routes bounded work to external AI CLI tools (Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, and Qwen Code) while the current Developer retains final responsibility. Covers read-only investigation, drift-audit, and code-review delegates, and read-write implementation delegates that create and edit files inside…
Skill Claude CodeCodex
Supervise autonomous execution of a frozen implementation plan (Mode B) - run one slice at a time in a fresh session, enforce the mechanical floor, assess every slice from repository evidence, commission independent review where risk warrants it, and stop for a human when the plan or the floor requires one.
Skill Claude CodeCodex
Create a concise evidence-backed written synthesis when the user explicitly asks for a report, assessment, investigation summary, comparison memo, status note, or final summary. Do not use for implementation planning, scoped implementation receipts, drift audits, code reviews, handoffs, PM run summaries, or commit…
Skill Claude CodeCodex
Implement one frozen acceptance slice from an implementation-plan without expanding scope. Use only when the user explicitly asks for this skill or provides an implementation-plan slice to execute. After this skill, the user should explicitly call drift-audit for authorization review.