agentleFS
hushh-labshushh-labs/hushh-research/.agents/skills/comms-community/SKILL.md

comms-community

Skill40 starsChanged 4 days agoApache-2.0View on GitHub

Use when drafting short community-facing replies for Discord, WhatsApp, or public chat about hushh-research, Kai, PKM, consent architecture, mobile/native tradeoffs, privacy boundaries, roadmap questions, or repo-based technical Q&A, including per-channel message formatting and length budgets.

The file

Show its rules and text

We split this skill into the rules it gives an agent

26 rules. About 0 are common, 16 are distinctive, and 10 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈16

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 10

    Very specific

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

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

    Avoid overlap with docs-governance and repo-context.

    line 12
  2. Distinctive:

    Read only the minimum repo docs needed to answer that exact question.

    line 53
  3. Distinctive:

    Start with the direct answer and separate shipped behavior from future direction.

    line 54
  4. Distinctive:

    if the premise is not grounded, say that directly before suggesting any fix

    line 59
  5. Distinctive:

    When the user asks for a Discord-formatted post, announcement, cinematic cadence, launch note, or channel message, use native Discord …

    line 62
  6. Very specific:

    use #, ##, or ### headings only when the channel message benefits from a clear top-level scan point.

    line 63
  7. Very specific:

    use -# subtext sparingly for one quiet context line, not as body copy

    line 64
  8. Distinctive:

    use short line breaks for cadence

    line 66
  9. Very specific:

    use > for one defining sentence or motto, and >>> only for a deliberate multi-line quote block

    line 67
  10. Distinctive:

    use bullets for scan paths, workflow steps, and proof points.

    line 68
  11. Distinctive:

    use inline code for repo paths, commands, lanes, branches, IDs, and exact file names

    line 69
  12. Distinctive:

    use fenced code blocks only for commands or copy/paste snippets, not for normal prose

    line 70
  13. Distinctive:

    use masked links when the URL is useful but visually noisy, and expose critical repo paths as inline code when teammates need to find local …

    line 71
  14. Distinctive:

    avoid spoilers, mentions, and role pings unless the user explicitly asks for them

    line 72
  15. Very specific:

    treat 2000 characters as the official hard message-content limit from Discord's developer docs

    line 74
  16. Very specific:

    use 1900 characters as the default safe copy budget so labels, edits, and pasted formatting do not push the message over the hard limit

    line 75
  17. Distinctive:

    before finalizing any Discord post, count the drafted message.

    line 76
  18. Distinctive:

    do not rely on Nitro/client-specific longer-message behavior unless the user explicitly asks for that target

    line 77
  19. Distinctive:

    prefer splitting at paragraph or section boundaries.

    line 79
  20. Very specific:

    use .agents/skills/comms-community/scripts/discord_chunk.py for deterministic chunking when the message is long or close to the limit

    line 80
  21. Very specific:

    When the user asks for WhatsApp formatting, a WhatsApp draft, or a mobile-chat paste, follow references/whatsapp-formatting.md.

    line 89
  22. Distinctive:

    If the question asks for the current architecture doc, cite the maintained top-level doc first and only then mention narrower subsystem …

    line 93
  23. Distinctive:

    Do not invent certainty from a vague teammate report.

    line 94
  24. Very specific:

    If the work becomes docs-home governance, use docs-governance.

    line 98
  25. Very specific:

    If the question cannot be answered cleanly without first mapping the repo or choosing the right owner family, start with repo-context.

    line 99
  26. Distinctive:

    If the task stops being public communication and becomes product or operational work, route to the correct owner skill.

    line 100

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Hussh Comms Community Skill

Purpose and Trigger

  • Primary scope: comms-community-intake
  • Trigger on Discord or public-chat reply drafting where the answer must be grounded in current repo docs and shipped boundaries.
  • Avoid overlap with docs-governance and repo-context.

Coverage and Ownership

  • Role: owner
  • Owner family: comms-community

Owned repo surfaces:

  1. .agents/skills/comms-community

Non-owned surfaces:

  1. docs-governance
  2. repo-context

Do Use

  1. Drafting concise public replies about shipped architecture, trust boundaries, roadmap boundaries, or repo-backed technical answers.
  2. Distinguishing clearly between current behavior and future direction.
  3. Selecting only the smallest set of evidence-bearing docs needed for the answer.
  4. Drafting repo-backed internal Q&A replies where the question may reference files, test failures, or implementation concerns that need verification before answering.
  5. Drafting Discord-ready or WhatsApp-ready messages with platform-native formatting and compact scan bullets.

Do Not Use

  1. Internal docs restructuring or repo-governance work.
  2. Product implementation, debugging, or operational workflows.
  3. Broad repo-orientation requests that should begin with repo-context.

Read First

  1. .agents/skills/comms-community/references/reply-rules.md
  2. .agents/skills/comms-community/references/whatsapp-formatting.md when drafting WhatsApp/mobile-chat copy
  3. .agents/skills/codex-skill-authoring/references/truth-first-operating-kernel.md
  4. docs/architecture/identity/README.md
  5. consent-protocol/docs/reference/developer-api.md

Workflow

  1. Infer the real architectural question before drafting the reply.
  2. Read only the minimum repo docs needed to answer that exact question.
  3. Start with the direct answer and separate shipped behavior from future direction.
  4. For repo-backed Q&A, verify the premise first:
    • confirm the referenced file, module, or test surface exists in the current tree
    • confirm the reported concern is actually visible in the current repo state when feasible
    • classify material claims with the truth-first labels before drafting
    • if the premise is not grounded, say that directly before suggesting any fix
  5. For material founder-language, product-direction, One/Kai/Nav, PCHP/BYOA, PKM/World Model, or roadmap replies, use the Founder Wiki North-Star Probe as a local evidence lane while keeping public replies public-safe. Private wiki evidence must not be cited or exposed unless the user explicitly asks for a private/internal draft; repo/wiki disagreement is current_state_vs_north_star_drift.
  6. For drafted reply/Q&A requests, default to one sendable reply: short, direct, evidence-backed, and sufficient to move the discussion forward. Add a second version only when the user explicitly asks for an alternative tone, more detail, or sharper correction.
  7. When the user asks for a Discord-formatted post, announcement, cinematic cadence, launch note, or channel message, use native Discord formatting deliberately:
    • use #, ##, or ### headings only when the channel message benefits from a clear top-level scan point; include a space after the heading marker
    • use -# subtext sparingly for one quiet context line, not as body copy
    • bold only the anchor phrase, decision, or headline; avoid bolding whole paragraphs
    • use short line breaks for cadence
    • use > for one defining sentence or motto, and >>> only for a deliberate multi-line quote block
    • use bullets for scan paths, workflow steps, and proof points; avoid Markdown tables because they are noisy in Discord chat
    • use inline code for repo paths, commands, lanes, branches, IDs, and exact file names
    • use fenced code blocks only for commands or copy/paste snippets, not for normal prose
    • use masked links when the URL is useful but visually noisy, and expose critical repo paths as inline code when teammates need to find local files
    • avoid spoilers, mentions, and role pings unless the user explicitly asks for them
  8. Discord message length guard:
    • treat 2000 characters as the official hard message-content limit from Discord's developer docs
    • use 1900 characters as the default safe copy budget so labels, edits, and pasted formatting do not push the message over the hard limit
    • before finalizing any Discord post, count the drafted message; if it exceeds the safe budget, split it into copy-ready batches
    • do not rely on Nitro/client-specific longer-message behavior unless the user explicitly asks for that target
    • for long posts, output Discord copy batch 1/N, 2/N, etc. with each batch inside its own fenced block so the user can copy one Discord-safe message at a time
    • prefer splitting at paragraph or section boundaries; if one paragraph exceeds the limit, split at a word boundary
    • use .agents/skills/comms-community/scripts/discord_chunk.py for deterministic chunking when the message is long or close to the limit
  9. For Discord posts that explain a Codex skill, use this structure by default:
    • headline: what was created or changed
    • one-line thesis in a block quote
    • What it does: 3-5 bullets focused on outcomes
    • How the flow works: compact numbered or bulleted flow from command/report to review action
    • Why it matters: trust, speed, duplication control, maintainer leverage
    • Where to look: local path, ignored/generated status, and the canonical skill/script surface
    • How to use it next: one concrete next command or workflow
  10. When the user asks for WhatsApp formatting, a WhatsApp draft, or a mobile-chat paste, follow references/whatsapp-formatting.md.
  11. Choose evidence format by audience:
  • public/community and internal shareable Q&A: prefer canonical GitHub markdown doc links on main, not repo-relative paths
  • internal repo-debug Q&A: file links or GitHub issue/PR links are allowed when they directly prove the point
  1. If the question asks for the current architecture doc, cite the maintained top-level doc first and only then mention narrower subsystem docs.
  2. Do not invent certainty from a vague teammate report. If the concern is branch-local or not present in the current tree, say so and ask for the exact path, log, or PR.

Handoff Rules

  1. If the work becomes docs-home governance, use docs-governance.
  2. If the question cannot be answered cleanly without first mapping the repo or choosing the right owner family, start with repo-context.
  3. If the task stops being public communication and becomes product or operational work, route to the correct owner skill.

Required Checks

./bin/hushh docs verify
python3 .agents/skills/comms-community/scripts/discord_chunk.py --self-test

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.