Agent
Delegate a task to the agy CLI and report back what it produced. Use when the user asks to run something with agy, or wants a task handled by a non-Claude model such as Gemini or GPT-OSS. BEFORE dispatching, if the user did not name a model, run agy models and ask them to pick one with AskUserQuestion. List the CLI's…
Agent
Verify that work claimed done actually runs and matches the spec. Use when a task is marked complete but unproven, the app should work end-to-end but doesn't, or a summary looks too clean.
Agent
Delegate a task to a separate claude CLI session and report back what it produced. Use when the user wants work done in a fresh Claude session with its own context. BEFORE dispatching, if the user did not name a model, ask them to pick one with AskUserQuestion. List the model value from /.claude/settings.json first…
Agent
Delegate a task to the OpenAI Codex CLI and report back what it produced. Use when the user asks to run something with codex. BEFORE dispatching, if the user did not name a model, ask them to pick one with AskUserQuestion. List the model value from /.codex/config.toml first, labelled (default). Pass the choice in the…
Agent
Delegate a task to the GitHub Copilot CLI and report back what it produced. Use when the user asks to run something with copilot. BEFORE dispatching, if the user did not name a model, ask them to pick one with AskUserQuestion. List the default model from /.copilot/config.json first, labelled (default); if it is not…
Agent
Delegate a task to the opencode CLI and report back what it produced. Use when the user asks to run something with opencode. BEFORE dispatching, if the user did not name a model, run opencode models and ask them to pick one with AskUserQuestion. List the model value from /.config/opencode/opencode.json first, labelled…
Skill Claude CodeCodex
Run the project's .feature files and report each scenario as passed, failed, pending, blocked, or unwired, with captured output as evidence and a suggested fix under every failure.
Skill Claude CodeCodex
Diff the current requirements against the existing .feature files and report the drift.
Skill Claude CodeCodex
Turn requirements into domain-driven Gherkin/Cucumber BDD scenarios.
Skill Claude CodeCodex
Write the step definitions that make the project's .feature files runnable, in whichever BDD framework the project uses.
Skill Claude CodeCodex
Review the architecture behind a branch, PR, MR, CL, commit range, folder, or whole repo: what the change was trying to do, diagrams where they earn their place, what needs attention, and a better shape for each.
Skill Claude CodeCodex
Interactively guides the user through generating a comprehensive system design document following a 7-step roadmap.
Skill Claude CodeCodex
Go (Golang) development guide. Use for any Go coding, review, debugging, or setup task — writing code, style and naming, errors, concurrency, context, testing, benchmarks, performance, security, databases, gRPC, GraphQL, CLI tools, dependency injection, linting, CI, observability, refactoring, project layout, and the…
Skill Claude CodeCodex
Read the paths you name against the clean code rules, rank each break by the smell it causes, and report it with a fix diff.
Skill Claude CodeCodex
Run the test suite with coverage on, rank the uncovered gaps by risk, and report per module and per file with a suggested test for each gap.
Skill Claude CodeCodex
The front door to the seven development dimensions: read whichever reports are on disk, say which dimensions this project should cover, cross-read them, and give one graded verdict on the code as written and tested.
Skill Claude CodeCodex
Test a running API against its own contract — OpenAPI, GraphQL, protobuf, or AsyncAPI — from the outside, and report every drift between what the contract promises and what the API does.
Skill Claude CodeCodex
Score every function with the CRAP metric — cyclomatic complexity crossed with test coverage — by generating one script that computes it, proving the script on fixed vectors, and reporting each failing function with the exact coverage it needs to pass.
Skill Claude CodeCodex
Find structurally duplicated functions across a codebase with a deterministic clone finder, separate the real duplication from code that only rhymes using git co-change history, and report each clone family with the knowledge it holds.
Skill Claude CodeCodex
Break the environment under a running system on purpose, measure how far the steady state moved and whether it came back, and report every weakness with the resilience mechanism that was missing.
Skill Claude CodeCodex
Write tests that run against the real database, broker, and other collaborators, prove each one red before green and the suite hermetic, and report which seams are covered and which are still unproven.
Skill Claude CodeCodex
Break the source on purpose, find the mutants the tests fail to catch, and report each survivor with the behaviour nobody is checking.
Skill Claude CodeCodex
The front door to the five release dimensions: read whichever reports are on disk, say which dimensions this system should cover, cross-read them, and give one graded verdict on whether the running system is ready to carry real traffic.
Skill Claude CodeCodex
Scan the code, its dependencies, its history, and the running app for security flaws, rank each one by whether an attacker can actually reach it, map the set to OWASP, and report every finding with a fix.