Run make targets and spawn fix sub-agents for quality checks.
npx skills add https://github.com/multicam/ikigai-rev --skill harness-multicam
Please help me install this Agent Skill. Skill: harness Source: https://github.com/multicam/ikigai-rev/tree/main/.claude-i/library/harness Command: npx skills add https://github.com/multicam/ikigai-rev --skill harness-multicam
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.
The Harness supports running structured quality checks, linking to fix scripts, and maintaining a centralized history to guide future attempts.
Run a harness to execute quality checks and spawn fix sub-agents automatically.
High-intent search queries and answers about installing and using this skill.
You automate quality checks by running a harness that executes a make target and spawns fix sub-agents to address failures, committing on success and reverting on exhaustion.
Escalation progresses from sonnet:think to opus:think to opus:ultrathink, allowing higher-level models to leverage historical learning from history.md to avoid repeating failed approaches.
The pattern uses .claude/harness/<name>/run paired with fix.prompt.md to enable reproducible quality gates by linking structured command patterns with versioned run scripts.
Yes, it applies directly to software projects requiring structured check pipelines, escalation pathways, and historical learning, satisfying metadata frontmatter and versioned run scripts for reproducible quality gates.
Each harness maintains a truncated per-file history.md for cross-attempt learning, where agents append summaries after each attempt so higher-level escalation models avoid repeating failed approaches.
xcrrrDefine evaluation goals, rubrics, and CI/CD pipelines for LLM systems.
loujcAutomate end-to-end idea discovery from broad topics to validated reports.
evolv3aiCreate and configure custom sub-agents in Claude Code via YAML.
krzemienskiExecute end-to-end validation journeys with per-attempt evidence and automated fix attempts.
