跳到正文
cccskills
免费在 GitHub 公开

factual-communication

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.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(2)

  • SKILL.md3.4 KB
  • README.md609 B

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

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 Ivan

Fact-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.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Use when the user requests integration testing, feature validation, or test plan execution

中文简介正在准备中。当前显示原文说明。

av/skills202026年10月21日 更新

anneal

免费

Use when the user wants to systematically fix AI code slop — duplicated logic, over-engineering, silent error swallowing, convention drift, cargo-cult patterns, and other LLM-introduced architectural decay — over a specified duration

中文简介正在准备中。当前显示原文说明。

av/skills202026年10月21日 更新

Produce a researched long-form article from a topic prompt via an orchestrated pipeline - research agent (first-person sources, working-definition gate), narrative-architecture outline, writer/cold-reviewer loop with an explicit ACCEPT/REVISE verdict contract, then a catalog-deslop pass with a regression gate. The orchestrator dispatches subagents only; the writer never judges its own draft. Use when the user says "article factory", "write an article about X", "run the article pipeline", or asks for a researched long-form piece produced end-to-end. For essays and micro posts in the user's own voice without a research stage, use the prose skill instead.

中文简介正在准备中。当前显示原文说明。

av/skills202026年10月21日 更新

Runs autonomous keep/discard experiments on a codebase to optimize a single metric for a fixed duration, in the style of karpathy/autoresearch. Use when the user says "autoresearch" (optionally with a focus, e.g. "autoresearch the optimizer"), asks to run experiments on a repo overnight, to hill-climb or optimize a metric autonomously, or points at a repo with a karpathy-style program.md.

中文简介正在准备中。当前显示原文说明。

av/skills202026年10月21日 更新

Create custom modules for [Harbor Boost](https://github.com/av/harbor/tree/main/boost), an optimizing LLM proxy. Use when building Python modules that intercept/transform LLM chat completions—reasoning chains, prompt injection, structured outputs, artifacts, or custom workflows. Triggers on requests to create Boost modules, extend LLM behavior via proxy, or implement chat completion middleware.

中文简介正在准备中。当前显示原文说明。

av/skills202026年10月21日 更新

bugbash

免费

Systematically explore and test any software project (CLI, API, Backend, Library, etc.) to find bugs, usability issues, and edge cases. Produces a structured report with full reproduction evidence (exact commands, inputs, logs, and tracebacks) for every issue.

中文简介正在准备中。当前显示原文说明。

av/skills202026年10月21日 更新

查看 av 的全部技能

报告此技能的问题