agentleFS
aaziraniaazirani/flutter_agentify/.agents/skills/agent-workflow/SKILL.md

agent-workflow

Skill0 starsChanged 4 days agoMITView on GitHub

How an agent works in Flutter Agentify, covering when to go ahead, when to stop and ask, how to record assumptions, how to leave a handoff note and how to end with a status line and change summary. Use when starting a task, hitting an unclear requirement, pausing before finishing or reporting results.

21 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use agent-agent, agent-analyze-code-quality and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 17 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Don't load this for which files you may edit.

    line 8
  2. Distinctive:

    Go ahead: when the request names the behaviour, the change stays inside the layers in AGENTS.md, and it is easy to undo.

    line 20
  3. Distinctive:

    Stop and ask: when:

    line 22
  4. Distinctive:

    Ask once, with the options and your pick: "A or B?

    line 29
  5. Distinctive:

    Assumptions: When you proceed on a guess, put it in your notes as one line each: "Assumed: items keep their order by creation time." Repeat …

    line 31
  6. Very specific:

    If you learned something the next session should know, follow capture-learnings.

    line 41
  7. Very specific:

    No automated check covers this behaviour.

    line 54

“Common” is measured: kept by at least 6 in 10 of the 17 edited versions of this file we could read. “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

Agent workflow

Don't load this for which files you may edit; see edit-boundaries.

Rule

Go ahead when the task is clear and the change is reversible. Stop and ask when it is not. Write down every assumption you make. End with a status line and a short change summary that says only what you ran.

Why

A person reading your result cannot see what you guessed. Asking one precise question early costs less than undoing a wrong feature. A report that claims a check you did not run is worse than no report. AGENTS.md says work is done only when tool/check.sh ends with All checks passed.

How

Go ahead when the request names the behaviour, the change stays inside the layers in AGENTS.md, and it is easy to undo. Pick the repo's own pattern and keep going.

Stop and ask when:

  • two readings of the request would produce different features or different data;
  • it needs a new dependency, platform configuration, CI change or a change to analysis_options.yaml (see edit-boundaries);
  • it deletes or rewrites the person's data, or changes a database table beyond adding tables or nullable columns;
  • the right action, route or screen does not exist and the scope is large;
  • a check fails and fixing it would mean raising a baseline or adding an ignore.

Ask once, with the options and your pick: "A or B? I will do A unless you say otherwise." Then continue with what does not depend on the answer.

Assumptions. When you proceed on a guess, put it in your notes as one line each: "Assumed: items keep their order by creation time." Repeat them in the final summary.

Handoff note. If you stop before finishing, leave a note another agent can continue from, in your reply (not a new file): the goal; what is done, with file paths; what is left; the exact commands to resume (for example tool/check.sh); what failed and the message; and open questions.

Ending. Finish with:

  1. A status line: Done, Blocked on <question> or Partly done: <what is left>.
  2. What tool/check.sh printed, or which steps you ran and which you did not.
  3. A change summary: each file or area changed, one line each, grouped by layer, plus any new action, route or translation key.
  4. The assumptions.

If you learned something the next session should know, follow capture-learnings.

Anti-patterns

  • Starting a large feature while the request is still ambiguous.
  • Asking several questions at once, or one you could answer by reading the code or the docs (see read-the-docs).
  • Saying "tests pass" after running only one test file.
  • Burying an assumption in the code instead of the summary.
  • A handoff note that says "mostly done" with no paths or commands.
  • Hiding a failing gate step instead of reporting it.

Verification

No automated check covers this behaviour. The gate (tool/check.sh) verifies the code only; the status line and summary are yours to get right. A reviewer can compare your summary with the files changed and with the gate's output.

Related skills

  • edit-boundaries — what needs a question first.
  • verify — running the gate and quoting its result.
  • capture-learnings — saving a correction or preference.
  • read-the-docs — answers you can find before asking.
  • adding-a-feature — the steps for a typical feature.

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.