Implement GitOps workflows with ArgoCD and Flux for automated, declarative Kubernetes...
npx skills add cordsjon/20_agentflow --skill "sh:troubleshoot"
Install specific skill from multi-skill repository
# Description
Issue resolution for builds, deploys, CI/CD, and system infrastructure
# SKILL.md
name: sh:troubleshoot
description: "Issue resolution for builds, deploys, CI/CD, and system infrastructure"
Troubleshooting
Diagnose and resolve operational issues: build failures, deployment problems, CI/CD pipeline errors,
dependency conflicts, and infrastructure configuration.
Broader than /sh:debug -- covers the full operational stack, not just code bugs.
When to Use
- Build failures (compilation, bundling, linking)
- Deployment problems (containers, services, environments)
- CI/CD pipeline errors (actions, hooks, scripts)
- Dependency conflicts (version mismatches, missing packages)
- Environment/configuration issues (env vars, paths, permissions)
Process
1. Classify Issue Type
| Type | Signals |
|---|---|
| Build | Compilation errors, missing modules, config parse failures |
| Deploy | Service won't start, port conflicts, permission denied |
| CI/CD | Pipeline step fails, secret not found, artifact missing |
| Dependency | Version conflict, peer dep warning, lockfile mismatch |
| Environment | Works locally but not in CI, missing env var, wrong path |
2. Gather State
- Read error output/logs completely
- Check environment: OS, runtime versions, env vars
- Review recent changes: git log, config diffs, dependency updates
- Verify prerequisites: required services, ports, credentials
3. Isolate Layer
For multi-layer issues, test each boundary:
Source -> Build -> Artifact -> Deploy -> Runtime -> Network
^ ^
Which layer fails? Which layer fails?
Add diagnostic output at each layer until you find where it breaks.
4. Diagnose
- Compare against known-working state (last green build, working branch)
- Check project-specific patterns (Makefile, docker-compose, CI config)
- Search error messages in project issues/docs
5. Resolve
- Apply targeted fix at the failing layer
- Verify the fix resolves the issue
- Confirm no regressions in adjacent layers
6. Report
## Diagnosis
- Issue: [what broke]
- Layer: [where it broke]
- Root cause: [why it broke]
## Resolution
- Fix applied: [what was changed]
- Verified by: [how confirmed]
## Prevention
- [Optional: how to prevent recurrence]
Boundaries
Will: Diagnose operational issues, apply safe fixes, verify resolution.
Will not: Modify production systems without confirmation, make architectural changes, ignore root cause.
Default behavior is diagnosis-only. Fixes require explicit user approval.
Next Step
After resolution, use /sh:test to run the test suite and confirm no regressions.
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.