factual-communication
Skill28 starsChanged 8 months agoMITView on GitHub
Restate content as a flat list of atomic facts in the style of av/facts fact sheets. Use when the user says "just facts", "facts only", "as facts", "give me the facts", or replies with a bare "facts" / "facts?" to a previous message — then the scope is that message's content.
In this repository25 other files it gives its agents.
- agent-integration-testingSkill
- annealSkill
- article-factorySkill
- autoresearchSkill
- boost-modulesSkill
- bugbashSkill
- bughuntSkill
- catalog-deslopSkill
- disciplineSkill
- ideateSkill
- improve-skillSkill
- make-videoSkill
- mosaic-writingSkill
- preact-buildless-frontendSkill
- pull-llamacpp-modelSkill
- run-llmsSkill
- sandcastleSkill
- superclaudeSkill
- timeboxed-iteratingSkill
- timeframe-researchSkill
- tinygradSkill
- turso-dbSkill
- use-coding-agentsSkill
- workgraphSkill
- workmachineSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 40
Keep attribution and provenance: when it matters: who said it, which source, which date — [source, YYYY-MM-DD] for volatile data.
- Distinctive: line 48
If the source is already minimal, say so in one line and don't pad.
≈ “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.
Files
factual-communication
Restate content as facts, in the style of av/facts fact sheets (github.com/av/facts). The output is a flat list of atomic claims: short enough to read in full, complete enough that nothing in scope is lost.
Trigger
- Explicit: "just facts", "facts only", "as facts", "give me the facts".
- Implicit: a bare reply "facts" (or "facts?") to a previous message → the scope is that previous message's content (yours or anyone's). If ambiguous, take the most recent substantive message in the thread.
- Scope = the content, not the conversation. Rehash what was said/reported, not meta-commentary about it.
Output format
One fact per line, - prefixed. Group with # headings only when there are 2+ distinct domains; skip headings for short outputs. No intro, no outro, no transitions, no "in summary". The fact list IS the message.
# billing
- MRR is $48.2k, up 6% MoM [stripe, 2026-08-25]
- 3 enterprise deals in contract stage
- Brain Restore Centers closed at $500/mo
# risks
- churn concentrated in workspaces without custom skills
- VIK-2058 unassigned until 07:20Z, now with IvanFact-writing rules
- Atomic — one fact = one truth. Split compound sentences. Each line stands alone with no dependence on neighbors.
- Minimal but sufficient — the test in both directions:
- Sufficient: a reader with only the fact list loses no decision-relevant information from the source. Numbers, names, dates, statuses, causal links, and caveats all survive.
- Minimal: remove any fact and something is lost; remove any word and precision drops. No hedging, no adjectives that don't constrain meaning, no restating a fact in different words.
- Declarative present tense where possible. "Deploy failed at 14:02" not "it appears the deploy may have failed".
- Concrete over abstract — keep the number, drop the characterization ("up 6%" not "healthy growth"). If the source only had the characterization, keep it but keep it short.
- Preserve epistemic status — distinguish known / claimed / uncertain. Mark non-verified claims inline:
- vendor claims fix ships Friday (unverified). Never upgrade a guess to a fact. - Keep attribution and provenance when it matters: who said it, which source, which date —
[source, YYYY-MM-DD]for volatile data. - Drop entirely: greetings, transitions, motivation/context the reader already has, recommendations unless they were part of the source content (then state them as facts:
- Viktor recommends X because Y). - Order by importance, not source order, unless the source is a sequence (timeline, steps) — then preserve order.
Sizing
- Bare "facts" reply to a message: usually 3–15 lines.
- Rehashing a document/report: sections + facts, still readable in <60 seconds.
- If the source is already minimal, say so in one line and don't pad.
Anti-patterns
- Bullet-ifying prose sentence-by-sentence (that's reformatting, not fact extraction).
- Losing a caveat or number to hit a line count.
- Adding analysis or facts not present in the source scope. If a critical gap exists, one final line:
- not covered in source: X.
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
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.

