Skill Claude CodeCodex
Senior-developer execution mode that returns the shortest SAFE working solution — a direct patch, command, or smallest runnable code — not frameworks, premature abstraction, multi-option essays, or future-proofing. Use whenever the user wants an implementation, bug fix, script, or config change, and ESPECIALLY on…
Skill Claude CodeCodex
Senior-developer execution mode for the shortest SAFE working solution: direct patch, command, or smallest runnable code. Use for implementation, bug fixes, scripts, config changes, and minimal-output signals like "ponytail", "lazy mode", "minimal", "shortest path", "just the fix", "don't overthink it", "no…
Skill Claude CodeCodex
Skill "v2.3.2" from cyfung1031/skills, covering ponytail, delta, shortest-safe-path ladder, hard boundaries and defaults.
Skill Claude CodeCodex
Use when starting a project, auditing an existing repo, planning a release, or writing project docs such as a spec, roadmap, release plan, engineering guide, changelog, or decision log. Provides a markdown-only, tool-neutral methodology for repo structure and living documentation.
Skill Claude CodeCodex
Generates source-grounded, evaluator-fit, first-shot-ready benchmark cases and evaluation suites with explicit claim mapping, source ledgers, oracle sealing, global differentiability, validation tiers, LLM evaluation controls, repeatable harnesses, audit ledgers, and release gates.
Skill Claude CodeCodex
Generates source-grounded, format-fit, first-shot-ready benchmark test cases that approximate non-synthesized real-world testing and produce globally differentiated, objectively scoreable evaluation suites.
Skill Claude CodeCodex
Generates source-grounded, format-fit, first-shot-ready benchmark cases and evaluation suites with explicit source-ledger, oracle-sealing, global-differentiability, validation-tier, LLM-evaluation, and release-gate controls.
Skill Claude CodeCodex
Generates source-grounded, evaluator-fit, first-shot-ready benchmark cases and evaluation suites with explicit claim mapping, source ledgers, oracle sealing, global differentiability, validation tiers, LLM evaluation controls, repeatable harnesses, audit ledgers, and release gates.
Skill Claude CodeCodex
Generate 16 strategically different variants of a source artifact, then iteratively synthesize, measure, and retest successors until no further measured improvement remains.
Skill Claude CodeCodex
Compare two raster images with deterministic pixel, residual, structural, edge, region, object-like component, artifact, and audit outputs. Use for visual regression, screenshot QA, and evidence-backed image-difference reports.