Diagnose and rewrite underperforming singularity-managed skills with iterative scoring validation.
npx skills add https://github.com/Shmayro/singularity-claude --skill repairing
Please help me install this Agent Skill. Skill: repairing Source: https://github.com/Shmayro/singularity-claude/tree/main/skills/repairing Command: npx skills add https://github.com/Shmayro/singularity-claude --skill repairing
Diagnoses why a singularity-managed skill underperforms or produces incorrect outputs and automatically rewrites it to improve reliability within the recursion loop.
Repair a failing skill by running the auto-repair routine on a known underperformer and review the repair history and score changes.
High-intent search queries and answers about installing and using this skill.
Automated skill repair diagnoses underperforming skills using history and telemetry, then rewrites SKILL.md sections to fix weaknesses while preserving high-scoring parts. It bumps version numbers and validates improvements through iterative testing.
Skill underperformance is diagnosed by analyzing telemetry data and scoring history to pinpoint low-scoring dimensions. The repair workflow targets these specific weaknesses through surgical rewrites rather than regenerating the entire skill.
Surgical skill rewriting preserves high-scoring sections of SKILL.md by targeting only identified weaknesses for modification. This approach maintains reliable functionality while correcting dimensions that caused outputs to fall below scoring thresholds.
Skill improvement validation runs iterative testing and scoring after rewriting SKILL.md and updating version numbers. The repair workflow confirms reliability gains by comparing new score records against previous thresholds to ensure measurable progress.
Automated diagnosis guides the repair of incorrect outputs by analyzing telemetry within the recursion loop. It applies a targeted rewrite workflow that updates SKILL.md, bumps versions, and re-tests to confirm reliability without external dependencies.
Run automated skill repair when average scores fall below a defined threshold or outputs are consistently incorrect. The workflow uses telemetry to diagnose failures, applies surgical rewrites, and validates results through re-testing and score record updates.
charlesjones-devRun type checks, linting, formatting, and tests across multiple project types.
mr-daedaliumMerge PRs, trigger deployment, and verify canary health in CI/CD pipelines.
thehivegroup-aiExecute a systematic test loop to identify, fix, and re-verify failures.
lawless-mOrchestrate test-fail-issue-fix loops with GitHub issues and Claude Code.
