agentleFS
rajasdevelrajasdevel/outcomebound/skills/explain-spec/SKILL.md

explain-spec

Skill0 starsChanged 15 days agoApache-2.0View on GitHub

Use when a person must act on a spec they did not write or says they do not follow it. A design walk-through qualifies; ask questions on it only when the person asks to test their understanding.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the spec and the code it points to, and teach only what they say.

    line 16
  2. Distinctive:

    When code falls short of a decided requirement, keep the requirement and record the gap separately.

    line 25
  3. Distinctive:

    Use the question and retry steps below only when the person asks to test their understanding.

    line 31
  4. Distinctive:

    When the spec does not answer the question, the gap is real: fix the spec in place, or record the row as open, in the same change.

    line 45

≈ “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

Explain a spec

You are done when the person can act on the spec: you have explained it to them and, where they asked to test their understanding, examined each misunderstanding that could change what they do; each real gap in the spec is fixed in it or recorded in it. You never report that the person fully understands the spec: you report what you explained, what they answered, and what changed in the spec.

Explain

Read the spec and the code it points to, and teach only what they say; say that anything else is not verified. Start from what the person says they already know and skip it. Give the outcome and who it is for, then each decision in dependency order with the alternative the spec rejected and the reason, then the edges: what happens when the input is empty, the call is repeated, a step fails, or the change is undone. Where order or flow is the point, add a diagram, in the form the surface shows. Write for the person in plain words, and in a file in the style the project sets for documents a person keeps; when a message does not land, say it again, shorter, in the project's own terms.

When code falls short of a decided requirement, keep the requirement and record the gap separately. Clarifying words must preserve the decision and its bounds; change the decision only within the authority the project grants.

Ask

Use the question and retry steps below only when the person asks to test their understanding. A request for an explanation alone requires no quiz or reply.

Draw the questions from the spec's rows, not from your own idea of what matters: for each decision with a rejected alternative, "why not" that alternative; for each assumed row, what is assumed and what would reverse it; for each edge, what happens. Ask first the questions whose answers would change what the person does. Ask for a prediction or an answer and wait for it; reading the explanation back is not understanding. Make the options of a multiple-choice question equal in length so the format gives no clue. Where trying a mechanism teaches more than reading about it, or the questions are many, the explanation and the questions can go in an explorable (the explorable skill), whose reply brings the answers back.

What a wrong answer means

When the spec does not answer the question, the gap is real: fix the spec in place, or record the row as open, in the same change. When the spec answers it and the person did not follow, explain that row again and ask again; where the row's words caused the misreading, fix the words in the same change.

With no person present

A run with nobody to answer cannot show understanding, so say that it is not verified. Put the questions, the rows most likely to be misread first, in your report, and in a file only where the task or the project names a place for it; go on with the rest of the work, since this skill blocks nothing.

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.