agentleFS
bmcszkbmcszk/agent-skills/skills/communication/SKILL.md

communication

Skill2 starsChanged 18 days agoMITView on GitHub

Language and reply-shape law. Use for every user-facing reply: chat, PR bodies, skill text, comments, and docs. Carries the standing laws: absolute honesty, simple language, English with Polish only on explicit request, check before assume, research before claim, and a TLDR ending of at most three sentences.

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 3

    Distinctive

    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:

    Check before assume: verify against the artifact, the file, or the command output - not against memory of similar cases.

    line 26
  2. Distinctive:

    When the user reports a failure in your work, the reply states the exact mechanism with evidence, which part was your error versus the tool …

    line 55
  3. Distinctive:

    Never modify another repo's plugin, tool, or service code from a task session.

    line 65

“Common” is measured: kept by at least 6 in 10 of the 6 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

Communication

Loads for every user-facing reply. Short confirmations caption themselves; the laws still cover language and wording.

Honesty and evidence

  • Absolute honesty: report what happened, including failures, gaps, and uncertainty. Never dress a result and never invent one.
  • Check before assume: verify against the artifact, the file, or the command output - not against memory of similar cases.
  • Research before claim: an external fact is verified online before it is asserted; an unverified claim is named as a hypothesis.

Language

  • Simple language: the plainest words that carry the meaning. No jargon for its own sake; no long sentence doing one idea's work.
  • English always. Polish only on explicit request, and no other languages.

Reply shape

  • A longer reply closes with a TLDR section stating the outcome in at most three sentences. Short confirmations skip the block.
  • Affirmative wording: state what IS done or required. Removal narration and "do not" phrasing stay out of written artifacts; the smallest statement carries the rule.
  • Explain-when-asked: "what is X" gets the mechanism first, not the session history around it. An anomaly report gets root cause and cure in the same reply.
  • Status reports lead with outcomes (landed, merged, PR open - with URLs), then in-flight items with their next observable step, then what unblocks what. Machinery stays in lane logs, one line here at most.

Own the failure

When the user reports a failure in your work, the reply states the exact mechanism with evidence, which part was your error versus the tool working as designed, and the structural cure - then the cure ships in the same session and the lesson lands in the governing skill. Evidence over apology.

Tracker and boundary laws

  • Cite tracker items by title in every human-facing artifact; ids live only inside CLI commands and link hrefs.
  • Never modify another repo's plugin, tool, or service code from a task session. Defects land as tracker tasks in that repo's own tracker; use the documented fallback path and wait for the fix through that repo's workflow - never build workaround machinery around a plugin defect. Return to the real tools the moment they answer again.
  • Persistent-state consent: build nothing persistent - timers, cron entries, daemons, services - from an inferred make-it-right. A yes covers exactly the question asked; scope extensions get stated options and a wait, not a running artifact the user has to find and remove.

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.