agentleFS
craigcossairtcraigcossairt/trellis/.claude/skills/trellis-survey/SKILL.md

trellis-survey

Skill5 starsChanged 12 days agoMITView on GitHub

Check this template's static wiring and setup - missing hooks, unreachable skills, unfinished context files. Runtime activation and trust require separate checks. Read-only. Use when asked "what do I have", "is everything set up", "what is actually running", "audit my setup", "did I break anything", "what did I delete".

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    You do not decide what is wrong. The script does: Read its output and relay it.

    line 19
  2. Very specific:

    If setup is unfinished, offer /setup.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

What this copy has, and how it is wired

bash bin/trellis-survey.sh

Exit 0 nothing to report, 1 there are findings, 2 could not run.

Exit 2 is not a clean bill of health. The survey could not complete its checks, which is different from finding no problems. Report it as what it is.

The rule

You do not decide what is wrong. The script does. Read its output and relay it. Do not go looking for additional problems by eye and mix them into the report - if the survey should catch something it does not, that is a change to the script, not a thing you improvise on top of it.

It is read-only, and that is the point

It writes nothing, deletes nothing and fixes nothing. Every finding is a decision for the person. Most of them have a perfectly good answer of "yes, I deleted that on purpose".

So: report first, offer second, and never fix as part of surveying. If they want something fixed, that is a separate action they ask for after seeing the list.

What the findings mean

HeadingWhat it meansWhy it matters
NOT RUNNINGThe harness or git calls something that is not there, or will skip itThis is the dangerous one. A missing hook does nothing and says nothing, so it looks exactly like a working hook
PRESENT BUT INERTA script is on disk and nothing invokes itHarmless, but it reads as protection that does not exist
HARNESS COVERAGEMissing or invalid routes in a retained adapterA skill with no Cursor or Codex router is invisible there; a router with no skill is a dangling pointer
SETUP UNFINISHEDFILL IN markers leftEvery one is a question the agent answers by guessing, in every session
TEMPLATE STATEVersion, and template files no longer presentUsually deliberate deletion. Listed so "I thought that was still here" has an answer

Lead with NOT RUNNING. Those are the ones where somebody believes they are protected and is not. The rest can be a short list underneath.

Limits to state rather than hide

  • Claude hooks are matched against settings.json by basename, not by parsing the JSON, because there is no guarantee jq exists on a fresh machine and a missing jq must not quietly turn the check into a pass. A hook wired under a different path with the same basename therefore reads as wired.
  • Codex uses bin/validate-codex.py to check JSON, event wiring, known script targets, router frontmatter, and coverage. It does not execute or interpret arbitrary shell commands. Python is needed only while a Codex adapter remains.
  • The survey checks static wiring, not runtime behavior or trust. A hook can pass this check and never run in an untrusted session. Follow SETUP.md's live checks with dummy files and inspect the hook verdict; a prose refusal from the agent is not evidence that the hook ran.

After the report

If setup is unfinished, offer /setup. If the template state shows a version, mention /trellis-sync will say what has changed upstream since. Do not run either one without being asked.

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.