From hw
Runs a structured preventive engineering audit using Intake-first, GQM, and ISO/IEC 25010. Use when users report poor experience, prepare refactoring, or need risk governance before delivery. Supports scope definition, independent audit worker separation, and tiered findings (Critical/Warning/Info).
How this skill is triggered — by the user, by Claude, or both
Slash command
/hw:auditThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
📌 输出语言规则:
📌 输出语言规则: 读取 config.yaml → output.language
使用此技能进行深度项目审计。/hw:audit 是风险治理 gate,不是普通代码质量建议清单;当用户遇到体验不好、准备重构、怀疑架构漂移或需要验收前风险判断时使用。
当 audit 用作验收或 worker-separation 证据时,在扫描开始前解决 Subagent/delegation 授权问题。audit worker 必须独立于实现被审计更改的 worker。如果授权缺失、被拒绝或不可用,停止或记录无法满足验收 gate 的降级审计;不要先在本地审计,然后声称缺少独立审查。
/hw:quality 或 /hw:optimize--scope <dir>--since <milestone>output.language 和 output.timezone。Critical、Warning 或 Info。.pipeline/audits/audit-NNN.md,使用 output.language,并按 references/completion-report-contract.md 写出完成说明:
无output.timezone 渲染报告时间戳。current.phase=lifecycle_audit。references/completion-report-contract.mdreferences/audit-spec.mdreferences/log-spec.mdSKILL.mdnpx claudepluginhub hypoxanthineovo/hypo-workflow --plugin hwIterative multi-agent code audit that detects project shape and risk signals, runs role-scoped reviewers, consolidates findings with a false-positive contract, and fixes critical/high issues in batches until clean.
Runs configurable codebase audits (health, evaluation, documentation) using parallel agent execution to produce intake docs for a pipeline.
Orchestrates extensible code quality audits: discovers dimensions, builds DAG for phased parallel execution via subagents, each in isolated context window.