harness
Automated quality check loops with escalation and fix sub-agents
When & Why to Use This Skill
Harness is a sophisticated automated quality assurance skill designed to create self-healing development workflows. It orchestrates automated fix loops by executing quality checks, spawning specialized sub-agents to resolve failures, and managing an intelligent escalation path across different AI model tiers. By maintaining a persistent history of attempts, it enables cross-attempt learning to ensure code stability, automatic commits upon success, and efficient error recovery without manual intervention.
Use Cases
- Automated Test Remediation: Automatically detect failing unit or integration tests and deploy sub-agents to diagnose and fix the underlying code issues.
- Continuous Code Quality Enforcement: Run complex linting or security scripts and have the agent automatically apply fixes to ensure the codebase meets predefined standards.
- Intelligent Debugging Escalation: Solve high-complexity bugs by automatically escalating tasks from standard models to high-reasoning models (like Opus Ultrathink) when initial fix attempts fail.
- Iterative Development Loops: Use the harness to maintain a 'green' build state by automatically reverting failed attempts and learning from previous errors stored in the history logs.
| name | harness |
|---|---|
| description | Automated quality check loops with escalation and fix sub-agents |
Harness
Automated fix loops. Each harness runs a make target, spawns sub-agents to fix failures, commits on success, reverts on exhaustion.
Pattern: .claude/harness/<name>/run + fix.prompt.md
Escalation: sonnet:think → opus:think → opus:ultrathink
History: Each harness maintains history.md for cross-attempt learning. Truncated per-file, accumulates across escalation. Agents append summaries after each attempt so higher-level models can avoid repeating failed approaches.
CLI: .claude/scripts/check-<name> symlinks to harness run scripts
Commands: /check-<name> runs the corresponding harness with --no-spinner
Quality: quality/run orchestrates all harnesses in sequence until stable
Running check-* Skills
When invoking check-* skills via Bash:
- Timeout: Use 60 minute timeout (
timeout: 3600000) - Foreground: Always run in foreground (never use
run_in_background) - Blocking: No output until completion—do not tail or monitor, just wait
- Output format: Structured JSON:
{"ok": true/false, "items": [...]}