agentleFS
seayniclabsseayniclabs/homelab-infra-mcp/skills/fix-and-deploy/SKILL.md

fix-and-deploy

Skill0 starsChanged 19 days agoMITView on GitHub

Fix a bug in a running homelab service, check the fix against the original symptom, apply it to the live container, proxy or DNS path, then check the live target again before reporting it fixed. Use when a fix must reach something already running, not only a commit or a local copy.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before you change anything, make the failure happen again:

    line 17
  2. Distinctive:

    Run the request that failed.

    line 21
  3. Very specific:

    Use the homelab-troubleshooting skill to find the cause when it is not clear.

    line 23
  4. Distinctive:

    Check the fix where you made it.

    line 31
  5. Distinctive:

    Every deploy change follows the dry run, go-ahead, approval flow.

    line 35
  6. Very specific:

    If the target is reachable from the internet, check it from outside your network.

    line 55
  7. Distinctive:

    If the target is public, the proof must be public too.

    line 72

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Fix and deploy

A fix counts as deployed when the live target shows the fix. A passing local test, a container that restarted, or a 200 from a health endpoint does not prove the reported bug is gone. Verify against the symptom the user reported, on the target where they saw it.

The sequence

1. Reproduce the original symptom

Before you change anything, make the failure happen again:

  • Record the exact target: the public URL, the container name, or the DNS name.
  • Record the exact symptom: the status code, the error text, or the behavior.
  • Run the request that failed. If you cannot reproduce it, you do not yet know what "fixed" means. Ask the user for the steps they took.

Use the homelab-troubleshooting skill to find the cause when it is not clear.

2. Fix the cause

Change one thing at a time. Write the cause in one sentence before you edit the config or image. Stop when the fix addresses that cause, not the symptom.

3. Check the fix before deploying

Check the fix where you made it. A local test or a dry run proves the change is valid. It does not prove the live service works. Call this the pre-deploy check, and report it as a separate step.

4. Deploy through Homelab

Every deploy change follows the dry run, go-ahead, approval flow. Pick the call that matches what changed:

What changedCallDry run first
Container image or configdocker_containers with restart, or pull then restart for a new image tagYes
Proxy host forward target or SSLproxy_hosts with updateYes
DNS record targetdns_records with updateYes
Certificate after a domain changeproxy_certificates with requestYes

Show each dry-run plan to the user. The person approves each change in the MCP client. If the client returns approval_unavailable, stop and explain the setting that the message names. Do not set it yourself.

5. Check the live target against the original symptom

Repeat step 1's request exactly. Check these first:

  • The container started after the deploy. Compare docker_containers with get: the state is running, and the start time is after the deploy. A container that did not restart still runs the old code.
  • diagnose_service with the domain. Every hop passes.

6. Check the public path from outside

If the target is reachable from the internet, check it from outside your network. Use a phone on cellular data, or a request from a host that is not on the LAN. Run curl -sI https://<domain> from that host and repeat the failing request.

A check from inside the network can pass while the public path still fails. This is the step that catches a stale DNS record, a missing certificate or a proxy rule that only applies locally.

7. Report

Report four facts:

  1. The original symptom, as the user reported it.
  2. The cause and the change.
  3. The pre-deploy check result.
  4. The live result, from outside the network when the service is public.

Say "fixed" only after step 6 passes. If step 6 does not pass, say which hop failed and what you checked.

The rule

If the target is public, the proof must be public too. A check that reaches the service only from the same machine proves less than the user's check.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.