[ Early development ] [ MIT License ]
A focused, open-source collection of AI-assisted troubleshooting skills for Hermes Agent — built to cut the time between “it’s failing” and “here’s the fix”.
Status: in early development. Both skills are planned, not released.
Independent community project, not affiliated with or endorsed by Nous Research.
Planned skills
Each skill is a folder with a SKILL.md of instructions that Hermes Agent loads when relevant. Small, modular, evidence-based.
Analyzes Jenkins and GitHub Actions failure logs, pinpoints the failing stage and the error that matters in the noise, explains likely root causes in plain language, and suggests fixes plus verification steps.
Explains common Kubernetes failure states — CrashLoopBackOff, ImagePullBackOff, Pending — from your kubectl describe output, events, and logs, then recommends the next most useful diagnostic command.
Principles
What the toolkit is designed around — one problem, one focused skill, one practical solution at a time.
Logs are noise. Skills are designed to find the failing stage and the line that actually matters.
Likely root causes in words engineers use — not opaque error codes left to interpretation.
Targeted remediation plus the commands or tests that confirm the fix actually worked.
Read-only diagnostics first. Destructive or production-affecting actions require explicit approval.
A consistent skill format makes it straightforward to add more DevOps and cloud skills.
Honest examples, reproducible scenarios, and contributions are welcome from day one.
How it works
You ask, or Hermes Agent selects the relevant skill for the problem at hand.
Paste logs, command output, or configuration. Sanitized — see the safety notes below.
Hermes uses its model and enabled tools, guided by the skill’s instructions.
The likely cause, clearly separated from inference, with recommended next steps.
You review every suggestion and choose what to run. The skill assists; you stay in control.
Requirements
# Requirements (any platform, planned)
# ─ Hermes Agent installed
# ─ A model provider configured in Hermes
# ─ Appropriate permissions for diagnostic tools
$ # install instructions: coming soon —
$ # being verified against current Hermes skill conventions# Requirements (any platform, planned)
# ─ Hermes Agent installed
# ─ A model provider configured in Hermes
# ─ Appropriate permissions for diagnostic tools
$ # install instructions: coming soon —
$ # platform-specific notes will land with the first release# Requirements (any platform, planned)
# ─ Hermes Agent installed
# ─ A model provider configured in Hermes
# ─ Appropriate permissions for diagnostic tools
$ # install instructions: coming soon —
$ # Windows-specific notes will land with the first releaseThe expected approach is copying a skill folder into your Hermes skills directory (~/.hermes/skills/). Exact commands will be documented and verified before the first release. Until then, see the Hermes Agent documentation.
This toolkit is meant to assist engineers, not replace operational judgment.
CI/CD logs and kubectl output can contain secrets. Never commit credentials, tokens, or kubeconfig files. Sanitize before sharing with any model provider.
Diagnosis should not change your system. Potentially destructive actions require your explicit approval, every time.
AI-generated diagnoses can be wrong. Always verify critical findings against the real system state before acting on them.
FAQ
No. The project is in early development and both skills are planned, not released. Everything described on this page is the design intent, not a finished feature.
The skills are written in the standard SKILL.md format for Hermes Agent. Adding support for other agent runtimes is not planned right now.
No. Independent community project, not affiliated with or endorsed by Nous Research. Hermes Agent is a product of Nous Research.
The toolkit is MIT-licensed and will be free. You still pay for your own model provider usage through Hermes Agent.
Suggest a specific DevOps problem worth solving, add sanitized troubleshooting examples, improve diagnostic instructions, or write tests for known failure scenarios. See CONTRIBUTING.md in the repository.