XRenSiu/claude-code-forge

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A debugging method that uses several agents to investigate competing explanations for a difficult software problem and challenge one another's findings.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A skill that creates a tailored DESIGN.md file from a product brief and a collection of existing design systems. The file records visual rules and the reasoning behind them.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A website design-analysis and cloning tool that extracts a structured profile of a site’s visual rules, such as colors, fonts, spacing, shapes, shadows, and effects. It can also use that profile to create a close code-based copy.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Orchestrator. Dispatches the 6 standalone review skills (code-reviewer / qa-reviewer / pm-reviewer / spec-drift-detector / spec-gaming-detector / meta-judge) against an implementation, runs the four-state ratchet (DONE / FIX / SPECDRIFT / GAMINGRISK), and persists every iteration to ratchet-log/iteration-NNN/. Owns…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Turn a fuzzy natural-language requirement into a machine-verifiable acceptance contract. Drafts EARS-format requirements (five sentence types), runs a strict clarify loop (only three question types allowed: ambiguity / missing edge / undefined term; 2-3 rounds, 3-5 questions per round, max 5 rounds total) to remove…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Review a code diff for bugs. Standalone skill — independent of any contract or pipeline; works on any PR / branch / ad-hoc snippet. Focus-driven (security / logic / perf / style / all) so you can spawn multiple focused passes in parallel instead of one "review everything" pass. Outputs structured findings (severity…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Synthesizer. Takes multiple independent review outputs (any subset of code-reviewer / qa-reviewer / pm-reviewer / spec-drift-detector / spec-gaming-detector) plus a pluggable rules source, and emits a single final-verdict.yaml with PASS / BLOCKMERGE / NEEDSHUMAN. The defining constraint: meta-judge does NOT re-review…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Verify code satisfies requirements. Standalone — works with EARS specs, Jira/ Linear tickets, PRDs, GitHub issues, or even PR descriptions. The skill's first step is requirement normalization: whatever format you pass in, it gets parsed into a flat bulleted list of requirements with stable IDs, then each is judged via…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Execute a test suite, collect evidence, classify failures, and output a release- readiness verdict. Standalone — independent of any contract; works with any test battery + thresholds config. The defining difference from LLM-as-judge review: this skill REALLY RUNS the tests, never asks an LLM "did this pass?". Every…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Code archaeologist. Detects factual divergence between a written spec (or doc, docstring, README, comment) and the actual code behavior. Crucially, this skill takes NO epistemological side — it does not assume the spec is right (that is pm-reviewer's stance), nor that the code is right. It only reports "these two…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Detective. Assumes the code author (human or AI) is gaming the contract. Standalone — works on any spec + artifact pair, not just donewhen.yaml. The defining stance: reversed default assumption ("find evidence of gaming, you fail if you say 'looks clean'"). Detects six RHD (Reward Hack Detection) patterns per…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Turn an EARS spec + donewhen.yaml contract into the full test pyramid: existence checks (ripgrep/tree-sitter), unit tests (example-based + property-based), integration tests (with testcontainers, no mocks), e2e (Playwright/Cypress/Appium/Maestro), and mutation-testing configuration as an anti-reward-hacking layer…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A team-based architecture design process in which several architects propose solutions, a critic challenges them, and an arbiter combines the results. An architecture is the high-level structure and technical choices of a system.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A requirements-writing process in which a product-focused agent and a technically skeptical agent debate a proposed feature for two rounds. It produces a requirements document after checking both the desired outcome and the codebase's technical limits.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A team-based review process that checks completed code and requirements from several viewpoints, including specification, code quality, security, and red-team attack testing. A final decision combines the findings.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A focused workflow for fixing a known software bug: investigate possible causes, apply a test-driven fix, verify it independently, and repeat if needed. TDD means writing a failing test first, then changing the code until the test passes.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A seven-stage development coordinator for Agent Teams that routes a request to feature development, bug investigation, or red-team review. It combines debate, parallel implementation, debugging, and review.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A team-based implementation workflow that assigns independent coding tasks to multiple agents while a lead coordinates shared files and progress. A separate checker regularly reviews the work.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A requirements-checking workflow that compares a written request with the current code and tests. It reports which parts are implemented, incomplete, or missing.

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Use when a codebase has NO shared ontology — when the team (and AI agents) lack a single agreed vocabulary of what the system's core objects, relationships, and constitutional rules ARE, and you want to reverse-engineer one from the existing code (and docs). Produces a Design Ontology Spec (DOS): a dos.yaml…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

Use when a Territory's □-class resident invariants are MISSING — when "what this block must always hold, on every Run" lives only in people's heads or in scars from past failures, not as written, traceable, verifiable rules. invariant-extract recovers them by ABDUCTION from the Territory's failure memory (the…

XRenSiu/claude-code-forge

Skill Claude CodeCodex

A product-definition method for describing how an experience or product world works, including its rules, meanings, and tests for correctness. It fills in missing decisions by asking the user questions.