OpenSpec Apply Change

SkillFiles & storage

Use this skill at Step 6 of the v2 SOP to implement one task from the change's tasks/ directory at a time. Each invocation should produce a code diff for exactly one task document, touching ≤ 2 files, with prompts pre-routed through mcp-model-router and any L3/L4 text pre-desensitized via phi-desensitize. Updates the task document checkbox upon completion. Chinese trigger examples: "执行任务", "OpenSpec apply", "Step 6", "实现 任务", "应用变更". Do NOT use to implement multiple tasks at once, do NOT use without an active MODEL_ALLOWLIST, do NOT use bypassing phi-desensitize for L3/L4 text. Success = one task's diff produced, ≤ 2 files touched, task checkbox flipped, all model calls routed through mcp-model-router.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the OpenSpec Apply Change skill

What this skill tells your AI

The instructions your AI receives, as published by charliehzm/medharness in .claude/skills/openspec-apply-change/SKILL.md and read by ahel’s review.

The implementation step. One task per invocation. Stay in scope.

Discipline

  • Read the task document; do exactly what it says.
  • Do not bundle unrelated cleanup.
  • Do not refactor adjacent code.
  • Do not add comments unless WHY is non-obvious.
  • If the task is wrong/incomplete, stop and route to prd-implementation-precheck instead.

Workflow

  1. Open the task document.
  2. Read the linked spec(s) for acceptance criteria.
  3. If any data-handling code involves L3/L4 → call phi-desensitize first; never paste raw values.
  4. Plan diff mentally; if > 2 files needed, re-decompose (call task-decomposition).
  5. Apply changes.
  6. Run quick local sanity (type-check, single test).
  7. Flip the task document checkbox to [x].
  8. Do NOT mark the OpenSpec change as done — that's Step 7 Verify.

Common failure modes

  1. Scope creep — "while I'm here, let me also fix..." . Mitigation: discipline; flag adjacent issues to spawn_task-style follow-up, not inline.
  2. Bigger diff than spec — finish the task with 4 files. Mitigation: pause at file 2; re-decompose.
  3. PHI in prompt — pasted real schema row including patient_id. Mitigation: hook + skill enforcement.
  4. Comments that re-describe code — adds noise. Mitigation: comments only where WHY is non-obvious.

Signals

GitHub stars
86
Forks
8
Last commit
Jun 2026
Advanced
Catalog kind
skill
Gateway key
openspec-apply-change-charliehzm
Source
github.com/charliehzm/medharness