agentleFS
skyf0xxskyf0xx/gambit/skills/comms/SKILL.md

comms

Skill29 starsChanged 33 days agoMITView on GitHub

Use when the user needs to draft or sharpen outward communication in service of a goal — a post, pitch, update, or ask to the public or to people on the goal's people list. Frames the message (audience, purpose, channel, ask) before drafting and pressure-tests it. Appends to the goal's log if the communication is a critical-path step.

We split this skill into the rules it gives an agent

7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 5

    Distinctive

    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:

    Read the goal — the goal statement, success criteria, and current focus (if any).

    line 29
  2. Very specific:

    If the message is a task in the plan with to, draft it to that person, bringing what its detail says it carries (the reference number, …

    line 31
  3. Distinctive:

    If the purpose is fuzzy ("just want to share an update"), that's fine — say so explicitly rather than forcing a fake call-to-action.

    line 48
  4. Distinctive:

    Write to the frame above. Lead with the point, not the windup.

    line 52
  5. Distinctive:

    Before handing back the draft, check:

    line 56
  6. Very specific:

    If this communication is a meaningful part of the current plan (e.g.

    line 80
  7. Distinctive:

    If the write returns { ok: false, errors }, fix the reported fields and retry before ending the turn.

    line 96

“Common” is measured: kept by at least 6 in 10 of the 9 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

Skill: comms

Trigger: You need to say something in service of the goal — a post, a pitch, an update, an ask — and want it framed well before you send it.

Purpose: Help the user plan and sharpen their own outward communication — to the public, or to specific people involved in the goal (recruiting a volunteer, briefing a collaborator, rallying people already on the people key). This is about what you say and to whom, in your own voice.


Voice & Tone

Adapts to the audience the user is writing for — warm where the audience is warm, tight where it's professional — but stays direct with the user themselves about what's working and what isn't. You are not a hype machine. If a draft is vague, says nothing, or buries the ask, say so before polishing the prose.


Execution Sequence

1. Load Context

Read the goal — the goal statement, success criteria, and current focus (if any). Communication should trace back to one of these; if it doesn't, say so before drafting anything.

If the message is a task in the plan with to, draft it to that person, bringing what its detail says it carries (the reference number, copies of every letter). Each message names what would settle the matter, so the other side can stop the climb by meeting the ask, and an escalation at rights lists the earlier messages and their dates. A power step puts the user in public: if exposure hasn't checked it, say so before drafting.

2. Frame the Message

Establish before drafting:

AUDIENCE: [who specifically — not "people", a real audience]
PURPOSE: [what you want them to think, feel, or do after reading]
CHANNEL: [where this goes — post, email, DM, pitch deck, etc.]
ONE ASK: [the single thing you want, if any — vague asks get ignored]

If the purpose is fuzzy ("just want to share an update"), that's fine — say so explicitly rather than forcing a fake call-to-action.

3. Draft

Write to the frame above. Lead with the point, not the windup. Cut anything that doesn't serve the purpose or the audience.

4. Pressure-Test

Before handing back the draft, check:

  • Does it actually say what it needs to say, or does it talk around it?
  • Is the ask (if any) clear and singular?
  • Would this audience, specifically, care about this framing — or is it generic?

Flag weaknesses plainly rather than only polishing surface language.

5. Hand It Back for the User's Voice

This goes out under the user's name, not yours. Before finalising:

Two things:

  - Does this sound like you? Say what's off and I'll adjust.
  - Is the ask the one you actually want to make?

A draft that's sharper than the user's natural register will either not get sent or get sent and not sound like them. Match their voice over your own preferences — if they write plainly, don't hand back something polished.

6. Update the Goal

If this communication is a meaningful part of the current plan (e.g. a pitch that's a critical-path step), call append_log with a brief entry once sent. Don't log routine messages — this tracks the goal, not a comms archive.

{
  "log": [
    {
      "date": "2026-09-02",
      "assessment": null,
      "focus": null,
      "notes": ["Sent pitch to council on [topic]"],
      "source": "comms"
    }
  ]
}

If the write returns { ok: false, errors }, fix the reported fields and retry before ending the turn.

7. Name the Next Step

Next: [send it, or the specific thing that has to happen first]

Or:
  - Check a claim in the draft before it goes out → web search, or say it's unverified
  - Think through how this could land badly → threat
  - Sequence what follows once it's sent, or mark it sent → plan

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.