agentleFS
CRELLIA-S-LCRELLIA-S-L/srs-dd/.claude/skills/srs-check/SKILL.md

srs-check

Skill0 starsChanged 32 days agoMITView on GitHub

Name and run the checks a finished change calls for — the specification checker, the grounds checker where the project keeps a register, the architecture checker where it keeps the layer, the tests the requirements it touched list in their own fields, and what a person has to look at where the method is not a test. Invoke when the user asks to check, verify or review work before committing, or when a change is finished and nothing has been run yet. Offers; never runs a suite unasked.

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. Very specific:

    Every requirement carries a verification method and the paths that verify it, and nobody reads them for this purpose — so the answer is …

    line 10
  2. Distinctive:

    Run anything the user did not approve.

    line 62

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
<!-- Copyright (c) 2026 CRELLIA S.L. — part of SRS-DD, distributed under the MIT License -->

Checking a finished change

The specification already knows which checks a change calls for. Every requirement carries a verification method and the paths that verify it, and nobody reads them for this purpose — so the answer is derived, not remembered.

Procedure

  1. Name what the change touched. The requirements it names, or, from the files it edited:

    python3 srs-dd/tools/srs_view.py --code <path>

    Name each of them as AGENTS.md asks the first time it appears — what this procedure reports is read by a person, and --cite prints the line. Say which paths you gave, not only which requirements came back: a path spelled wrong returns an empty list, and an empty list is also what a change governed by nothing looks like.

  2. Read what each of them asks for. python3 srs-dd/tools/srs_view.py <ID> prints the method and the tests field. Sort them into lists:

    • the checker — always, for every change to specs/: python3 srs-dd/tools/srs_check.py;
    • the architecture checker, where the project carries the layer — arch/arch-config.json beside specs/, what the shipped pipeline reads it by: python3 srs-dd/tools/srs_arch.py. It regenerates arch/90-map.md on the same terms as the dashboard below, and a change that moved a file between parts is exactly what it reports.
    • the grounds checker, where the project carries a register — grounds/grounds-config.json beside specs/, what the gate reads it by: python3 srs-dd/tools/srs_grounds.py. It regenerates grounds/90-dashboard.md, and the project's gate fails on a committed copy that no longer matches, so a change that moved a requirement or a record and did not run this hands back work that passes everything else and reddens the pipeline. Where there is no register the command is not there, and this line does not apply — say nothing about it;
    • the suites those requirements name in tests, and nothing else: a change to the viewer does not call for the installer's suite;
    • what a person has to look at, where the method is I or D, and the measurement to retake or the analysis to reread where it is A. Say it in words, naming the requirement and what to do: "this one is verified by inspection — open the page, follow a link from the dashboard, and see that it lands on the card".

    A requirement whose method is T with an empty tests field is a gap, not a check: say so rather than inventing something to run.

  3. Offer, and wait. Builds and test runs need the user's word each time — that is SRS-DD-ART-030 of srs-dd/method.md, not politeness. List the commands and ask. Where the project has checks of its own that the specification does not name, ask rather than guess.
  4. Run what was approved, all of it, and report per check: passed, failed with the output, or not run and why. Read that report once before sending it: a result is a plain statement, and nothing about the running of the checks belongs in it.
  5. After a repair, offer the whole set again. A fix is a change like any other: rerunning the check that was red says it is no longer red and nothing about what the fix broke beside it. The list is the one from step 2, unchanged — a repair does not narrow the change it belongs to — and it is offered, not run (SRS-DD-ART-030). Where the repair touched the specification, the checker is in that list too.
  6. Say what is left to eyes. The inspection list from step 2 does not disappear because the suites are green. A change verified by inspection and never inspected is unverified.

What this procedure does not do

Commit. Decide that a red check is acceptable. Run anything the user did not approve. Replace the project's own habits — where a project has a gate of its own, this names what the specification asks for on top of it.

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.