OpenSpec Apply Change
SkillFiles & storageUse 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.
No other account needed.
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-precheckinstead.
Workflow
- Open the task document.
- Read the linked spec(s) for acceptance criteria.
- If any data-handling code involves L3/L4 → call
phi-desensitizefirst; never paste raw values. - Plan diff mentally; if > 2 files needed, re-decompose (call
task-decomposition). - Apply changes.
- Run quick local sanity (type-check, single test).
- Flip the task document checkbox to
[x]. - Do NOT mark the OpenSpec change as done — that's Step 7 Verify.
Common failure modes
- Scope creep — "while I'm here, let me also fix..." . Mitigation: discipline; flag adjacent issues to
spawn_task-style follow-up, not inline. - Bigger diff than spec — finish the task with 4 files. Mitigation: pause at file 2; re-decompose.
- PHI in prompt — pasted real schema row including patient_id. Mitigation: hook + skill enforcement.
- 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