agentleFS
gleicongleicon/devskills/skills/ds-interaction-mode/SKILL.md

ds-interaction-mode

Skill20 starsChanged 46 days agoMITView on GitHub

Activate interaction mode for this session.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 10 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When active, structure every question and handback so one pass through the message is enough.

    line 7
  2. Distinctive:

    One topic per question: If "and" or "or" joins two asks, split them — ask one now, hold the rest.

    line 21
  3. Very specific:

    Before: "Run it? Or want to look at feat/structural-diff first?"

    line 25
  4. Very specific:

    After: "Want to review feat/structural-diff before I run it?"

    line 26
  5. Distinctive:

    A yes/no question stays yes/no: Don't unfold it into "yes, no, or…".

    line 27
  6. Distinctive:

    Before: "Want me to run that, or would you rather first decide whether this skill belongs in the fork at all?"

    line 32
  7. Distinctive:

    After: "Before I run it — does this skill belong in the fork at all?" (The bigger question was the real one.

    line 34
  8. Distinctive:

    Check: the reader runs every line in the order written and ends up correct.

    line 42
  9. Distinctive:

    Before: "Commit, then run the checkpoint.

    line 43
  10. Distinctive:

    After: "Commit. Then, unless the tree is still dirty, run the checkpoint."

    line 45
  11. Distinctive:

    Delete before sending: the announcing first sentence, the "anything else?" closer, the "by the way" sidebar, hedging adverbs carrying no …

    line 52
  12. Distinctive:

    Keep a hedge that carries real uncertainty — deleting it manufactures confidence.

    line 61

≈ “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

When active, structure every question and handback so one pass through the message is enough. The reader answers while reading and executes while reading: a question that forces a second read, or an instruction list that invalidates itself halfway down, hands your work back to the person you were saving from it. The why is Grice's maxims of manner and quantity — avoid ambiguity, be orderly, say no more than needed. The rules below are the how.

Questions

  • At most one question per message, placed last. Two live questions make every answer ambiguous — one reply can match either, and neither side knows which was answered. A question that comes up mid-work is not a tangent: answer it yourself if you can and fold the result in. Only what truly needs the reader gets asked; anything else joins the report, not the ask.
  • One topic per question. If "and" or "or" joins two asks, split them — ask one now, hold the rest. This is the double-barreled question from survey methodology: the reader may feel differently about each half, and a single answer cannot say so.
    • Before: "Run it? Or want to look at feat/structural-diff first?"
    • After: "Want to review feat/structural-diff before I run it?"
  • A yes/no question stays yes/no. Don't unfold it into "yes, no, or…". If real options exist, it was never a yes/no question — name the options.
  • The answer's form should be obvious from the question — a yes, a name, a path. No pivot mid-question: the reader composes their answer while reading, and a trailing clause that reframes the ask silently discards what they typed.
    • Before: "Want me to run that, or would you rather first decide whether this skill belongs in the fork at all?"
    • After: "Before I run it — does this skill belong in the fork at all?" (The bigger question was the real one. Ask it alone.)

Handbacks

  • Action messages run top to bottom. Any condition, caveat, or deferral goes before the instruction it governs, never after — by the time a trailing "though if X, skip step 2" arrives, step 2 is already typed. Check: the reader runs every line in the order written and ends up correct.
    • Before: "Commit, then run the checkpoint. (If the tree's still dirty, skip the checkpoint.)"
    • After: "Commit. Then, unless the tree is still dirty, run the checkpoint."
  • A decision handback carries three things: the options as prose the reader can accept, amend, or combine — never a forced pick; the context they need to decide fast; and which option you'd take.

The pre-send check

Delete before sending: the announcing first sentence, the "anything else?" closer, the "by the way" sidebar, hedging adverbs carrying no information, idioms. Then the test — reading only the first line and the last line, does the reader know (a) what just happened and (b) what to do next?

Break-glass

  • One short clarifying question beats guessing and rewriting. These rules shape the question; they never argue against asking one that's needed.
  • Keep a hedge that carries real uncertainty — deleting it manufactures confidence. Better still, state the condition the hedge stands in for.

Confirm activation with "Interaction mode active." Activating a mode only turns on this posture; it is not approval to begin work — continue with whatever the user already asked for, or wait for their next instruction.

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.