Communitygithub.com

rfdiosuao/enterprise-debt-remediation-skill

面向 Codex、Claude Code、OpenClaw 与 Hermes 的证据驱动技术债审计和修复 Skill

¿Qué es enterprise-debt-remediation-skill?

enterprise-debt-remediation-skill is a Claude Code agent skill that 面向 Codex、Claude Code、OpenClaw 与 Hermes 的证据驱动技术债审计和修复 Skill.

Compatible conClaude CodeCodex CLI~Cursor
npx skills add rfdiosuao/enterprise-debt-remediation-skill

Preguntar en tu IA favorita

Abre un nuevo chat con esta habilidad de agente ya precargada.

Documentación

HZ

Run the smallest useful debt pass. Default to audit when no mode is given. Treat this as the thin entrypoint for $enterprise-debt-remediation.

REQUIRED SUB-SKILL: Use qiaomu-goal-meta-skill when generating a debt-remediation /goal.

In Codex, use $hz goal ... to pass the goal mode to this skill. Some clients show only the single $hz skill entry and do not display goal as a separate selectable subcommand.

Environment Preflight

Detect the host OS before routing. If the host is Windows, apply the UTF-8 session bootstrap first, then continue with the chosen mode.

Modes

audit

Scan the project and return only the highest-value debt items with evidence, impact, priority, and validation.

plan

Turn the current findings into a short remediation plan with order, owner, and verification.

fix

Apply or draft the smallest safe changes for P0/P1 items only. Stop if scope, evidence, or safety is unclear.

goal

Audit the requested scope first, turn the findings into a prioritized debt ledger, then output a copy-ready Codex /goal for resolving that ledger. Do not start fixing in goal mode unless the user explicitly asks to execute the generated goal.

Treat "all debt" as all evidence-backed findings inside the requested scope. If areas were not inspected or evidence is weak, list them as blockers or pending-evidence items instead of claiming the audit is exhaustive.

Routing

Interpret /hz [mode] [scope] as:

  • mode: audit, plan, fix, or goal
  • scope: everything after mode

If mode is missing or unknown, assume audit and state the assumption. For complete audits, load $enterprise-debt-remediation and follow references/core-contract.md. If the host is Windows, ensure the UTF-8 session bootstrap has been applied before loading the core contract.

Route these requests to goal even if the literal mode is omitted: "find all debt then make a goal", "generate a goal to fix the debt", "debt remediation goal", audit-goal, solve-goal, or all-goal.

For goal, first run the complete debt-audit workflow from $enterprise-debt-remediation, then apply the /goal contract from qiaomu-goal-meta-skill:

  • Outcome: resolve the audited debt ledger in priority order.
  • Verification: use repo-discovered checks, runtime evidence, screenshots/logs when relevant, and per-item validation from the debt ledger.
  • Constraints: preserve public behavior, data shape, user data, secrets, default branches, and unrelated systems.
  • Boundaries: name allowed write paths and forbidden paths; keep production data, credentials, destructive operations, and unrelated rewrites out of scope.
  • Iteration policy: fix one priority batch at a time, rerun checks after meaningful changes, update the ledger, and switch evidence source after repeated failures.
  • Stop when: every confirmed debt item is fixed and verified, or explicitly deferred with owner/reason/residual risk, and final checks are reported.
  • Pause if: credentials, production data, paid services, destructive operations, legal/compliance decisions, unclear ownership, or scope expansion is required.

Output

For audit, plan, and fix, return only:

  • findings
  • next_step
  • blockers

For goal, return only:

  • findings
  • remediation_goal
  • blockers

Skills relacionados