vibeguard

Use this skill whenever VibeGuard hook latency, benchmark-action output, GitHub Actions bench-output artifacts, or a suspected performance regression needs investigation. It preserves the PR-to-CI-artifact workflow for comparing recent PRs, older anchors such as PR

Use when the user asks for a VibeGuard dashboard, hook health, stats, doctor output, metrics export, or local observability status from the Codex App plugin.

Use when the user asks to install, verify, inspect, or remove VibeGuard Codex hooks through the Codex App plugin.

Use when the user asks Codex to apply VibeGuard anti-hallucination rules, inspect VibeGuard observability, or route a task through VibeGuard preflight/review/check workflows.

AI-assisted development of anti-hallucination specifications. Check out the seven-layer defense architecture, quantitative indicators, execution templates and practical cases. Used for code review, task startup inspection, and weekly review.

Automate analysis, evaluation, design and optimization of target projects. Integrate VibeGuard as a baseline scan, the remediation process adheres to VibeGuard specifications, and a compliance check is run at the end. Support auto-run-agent autonomous execution.

Discover and deliver repository optimization work end to end: identify performance/reliability/maintainability/security/dx/cost optimization points, prioritize by impact-effort-risk, then execute fixes step by step with continuous testing and explicit commit policy (`final_only`, `per_step`, `milestone`). Default to `per_step` for implementation work so each feature is tested then committed before the next feature. Supports optional BDD (Given/When/Then).

Audit AGENTS.md / CLAUDE.md against the five high-leverage patterns (progressive disclosure, procedural workflows, decision tables, production code examples, domain rules with concrete alternatives). Reports per-pattern coverage, anti-patterns, and a prioritized fix list.

Shell 脚本中 awk 的 POSIX 兼容性指南。 Use when: 编写或审查包含 awk 的 shell 脚本， 尤其是需要 macOS + Linux 跨平台运行的场景。 触发词: awk, BSD awk, POSIX regex, [[:space:]], guard 脚本, 跨平台 shell

Assessment-driven development — Quantify code generation quality with pass@k / pass^k metrics, automatically scored by Grader.

Iterative retrieval — 4-stage loop (DISPATCH→EVALUATE→REFINE→LOOP) to pinpoint relevant information in the code base. Up to 3 rounds.

Strategic compression — Manual compression of contexts at logical boundaries rather than arbitrary automatic compression. Key decisions and constraints are preserved and intermediate exploration is discarded.

Post-hoc diagnosis of a failed agent trajectory. Classifies the first unrecoverable step into one of nine failure categories (plan adherence, hallucinated information, invalid tool call, misread tool output, intent–plan mismatch, under-specified intent, unsupported intent, guardrail trigger, system failure) and produces an evidence-backed root-cause report.

Execute coding tasks with a strict delivery workflow: build a full plan, implement one step at a time, run tests continuously, and commit by default after each step (`per_step`). Support explicit commit policy overrides (`final_only`, `milestone`) and optional BDD (Given/When/Then) when users ask for behavior-driven delivery or requirements are unclear.

Analyze repository-level duplicate/redundant design first, then build and execute a strict step-test-update plan in docs/plan. Use for deep architecture review, convergence planning, and traceable one-step-at-a-time delivery.

Use when the user asks to enter Plan mode, says /prompts:plan or /plan, or wants a structured execution plan written to plan/.