maicongambini/omp-harness-guide/harness-role-separation
Claude Code and agent skill by maicongambini/omp-harness-guide
- community
Checking that the SKILL.md file is still in the repository.
Separate planner, generator, and evaluator roles. Use when handling complex, UI-heavy, or subjective tasks. Do NOT use for scheduling parallel lanes (use orchestrate).
Install
npx vskill@latest install maicongambini/omp-harness-guide/harness-role-separationSKILL.md
The same SKILL.md was published earlier as maicongambini/omp-harness-guide/harness-role-separation.
Harness Role Separation
Use when a task is complex, has subjective outcomes, touches UI, or where self-review would miss defects.
When to Use
- Task involves UI/visual design decisions
- Task spans multiple files or subsystems
- Prior similar tasks had quality drift or missed requirements
- Task is high-risk (auth, data integrity, public API)
- Single-role approach produced incomplete output before
Role Definitions
Planner: Defines scope, writes sprint contract, defines AC, does NOT write implementation. Generator: Implements ONLY what is in sprint contract scope. No scope expansion. Evaluator: Judges against sprint contract AC using harness-evaluator-rubric. Does NOT self-approve.
In OpenCode, one agent can play all three roles — but sequentially, with explicit role transitions.
Workflow
1. Planner Phase
Fill docs/harness/sprint-contract.md before writing any code:
- Sprint Overview: what and why
- Scope In: explicit list of what will change
- Scope Out: explicit list of what will NOT change
- Roles: planner / generator / evaluator
- Acceptance Criteria: observable, runnable
- Verification Plan: commands + pass conditions
- Evidence Log: leave empty, fill during/after Execute
State: "Planner complete. Switching to Generator role."
2. Generator Phase
Implement ONLY sprint contract scope:
- If you discover adjacent work → add to
feature_list.jsonasnot_started. Do not implement now. - If scope must change → stop, update sprint contract, get approval, then continue.
State: "Generator complete. Switching to Evaluator role."
3. Evaluator Phase
Invoke harness-evaluator-rubric. Do not produce verdict in same response as implementation. Fill Evidence Log in sprint contract with actual command output. Record revision log entry if any rework was needed.
Revision Log Pattern
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingomp/agent/skills/harness-role-separation/SKILL.md
Popularity
- Repository stars
- 2
- 7-day trend score
- 5
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/maicongambini/omp-harness-guide/harness-role-separation)