agentleFS
alimohammadiamirhosseinalimohammadiamirhossein/deputy/skills/ds/SKILL.md

ds

Skill2 starsChanged 21 days agoMITView on GitHub

Hand every task in this session to Deputy (DeepSeek by default), in any mode, until the user says "/deputy:ds off". Deputy reads, writes and tests the code; you only write the brief, check the result and report. Use when the user types "/deputy:ds" or asks to let DeepSeek or Deputy do the work.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Brief, don't read: Do not open or edit files yourself first.

    line 18
  2. Very specific:

    Let Deputy find things: If you need to know where something is before you can write the brief, ask investigate instead of reading files …

    line 23
  3. Very specific:

    Delegate: Call delegate with the brief.

    line 25
  4. Distinctive:

    Check cheaply: Run the verify command and look at git diff --stat.

    line 27
  5. Very specific:

    Send fixes back: If a check fails, call delegate again with the failing output and what must change.

    line 29
  6. Distinctive:

    If this is empty or "on", switch on and wait for a task.

    line 42

≈ “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

Let Deputy do the whole task

The user wants Deputy to do the work, not you. This is their explicit request, so it applies in every mode, accuracy included. Your job is to brief, check and report, not to write the code.

This stays on for the rest of the session. Handle every later coding task in this conversation the same way, without the user typing the command again, until they say /deputy:ds off or ask you to stop using Deputy. When it is switched off, go back to the mode's normal rules and say so in one line. When it is switched on without a task, say in one line that Deputy will now handle the tasks in this session, and wait for the first one.

  1. Brief, don't read. Do not open or edit files yourself first. Turn the request into a brief for delegate: the goal, the files or folders in scope, constraints (what not to touch), and a command that verifies the result (tests or a build). Put facts Deputy cannot discover, such as decisions already made or error messages, in context. Never include secrets. If the request is too unclear to brief, ask the user one short question instead of guessing.
  2. Let Deputy find things. If you need to know where something is before you can write the brief, ask investigate instead of reading files yourself.
  3. Delegate. Call delegate with the brief. Split a large task into 2-4 self-contained steps and delegate them one after another.
  4. Check cheaply. Run the verify command and look at git diff --stat. Read the diff only where it is risky: deleted code, security, data handling, files outside the scope.
  5. Send fixes back. If a check fails, call delegate again with the failing output and what must change. Do not fix it yourself. After two failed attempts, stop and tell the user what is still wrong, and offer to finish it yourself.
  6. Report in a few lines: what Deputy changed, whether the checks pass, anything you could not verify, and Deputy's cost from its result.

Skip review and consult here unless the change is risky. They use the same cheap model that just wrote the code, and the user asked for speed.

What the user typed after the command

$ARGUMENTS

If this is empty or "on", switch on and wait for a task. If it is "off", switch off. Otherwise it is the first task: start on it now.

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.