Skill Claude CodeCodex
Interviews the project and generates a complete .claude/rules/ rule set that agents, new teammates, and handover recipients can immediately reference. Extends an existing CLAUDE.md if found; generates from scratch if not.
Agent
Analyzes API route files, middleware, response formats, and validation patterns. Used by init-rules skill during parallel codebase analysis. Only spawned when an API layer is detected.
Agent
Analyzes project directory structure, module boundaries, dependency graph, data flow, global state, and key dependencies. Used by init-rules skill during parallel codebase analysis.
Agent
Analyzes CI/CD workflows, Docker configuration, and deployment patterns. Used by init-rules skill during parallel codebase analysis. Only spawned when CI/CD config files are detected.
Agent
Analyzes source file naming conventions, import styles, error handling, async patterns, component patterns, and code style with frequency data. Used by init-rules skill during parallel codebase analysis.
Agent
Analyzes how major dependencies are used — direct vs wrapped, custom abstractions, and usage consistency. Used by init-rules skill during parallel codebase analysis.
Agent
Analyzes git history, commit message patterns, branch naming, and merge strategy. Used by init-rules skill during parallel codebase analysis.
Agent
Cross-validates findings from all analysis agents, detects conflicts, rates pattern consistency, and produces a unified analysis. Used by init-rules skill after parallel codebase analysis.
Agent
Analyzes test file structure, framework, patterns, boundaries, data patterns, and type distribution. Used by init-rules skill during parallel codebase analysis.
Agent
Validates generated rule files against the actual codebase. Reports match percentages and flags rules that conflict with existing code. Used by init-rules skill after rule files are generated.
Skill Claude CodeCodex
Converts academic papers into implementable code functions. Accepts a PDF path, arxiv URL, or search keyword — if a keyword is given, searches for relevant papers and lets the user choose. Extracts mathematical formulas, algorithms, and logic, then generates well-documented functions with paper references and…
Agent
Generates code functions from extracted paper elements. Follows dependency ordering, uses meaningful variable names, and includes paper references in docstrings. Used by paper2code skill in STEP 6.
Agent
Identifies implementable elements from a paper — both mathematical formulas AND domain knowledge (rules, classifications, causal relationships, thresholds). Builds a dependency DAG for correct code generation ordering. Used by paper2code skill in STEP 3. Does NOT re-read the paper.
Agent
Reads academic paper content from PDF or URL. Handles chunked reading for large PDFs (>20 pages). Extracts structure, formulas, algorithms, and notation. Used by paper2code skill in STEP 2.
Agent
Searches for academic papers across multiple sources (web search as primary, Semantic Scholar as secondary). Optimizes queries, deduplicates results, and pre-validates PDF accessibility. Used by paper2code skill in STEP 1.
Agent
Validates generated code against paper definitions. Checks variable coverage, dimensional consistency, operation correctness, and edge case handling. Static analysis only — does not execute code. Used by paper2code skill in STEP 7.