agentleFS
Sign inStart free
DeiblerDeibler/edmund-harness/skills/summarize/SKILL.md

summarize

Skill24 starsChanged 2 months agoMITView on GitHub

TL;DR a URL, YouTube video, or local PDF/file via the `summarize` CLI (Gemini Flash). Use when the user sends a link and asks what it's about, drops a YouTube URL and wants the gist, or has a PDF in the sandbox they haven't opened.

  • Reads credentials

93 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use openclaw-performance, plugin-sdk-impact and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • clawdbotSkill

    openclaw-performance

    Performance work in OpenClaw: a slow or heavy flow (latency, CPU, memory, SQL, RPC, renders), writing or reviewing a perf PR, or adding a cache. Count the work, delete waste first, cache last.

    Also in 7 projects that use this

  • clawdbotSkill

    plugin-sdk-impact

    Count Plugin SDK consumers across live ClawHub releases and bundled OpenClaw plugins before deprecation, removal, or behavior changes; supports public release-date filters.

    Also in 7 projects that use this

  • AtomicBot-aiSkill

    taskflow-inbox-triage

    Also in 6 projects that use this

  • hkyutongSkill

    node-connect

    Diagnose OpenClaw node connection and pairing failures for Android, iOS, and macOS companion apps. Use when QR/setup code/manual connect fails, local Wi-Fi works but VPS/tailnet does not, or errors mention pairing required, unauthorized, bootstrap token invalid or expired, gateway.bind, gateway.remote.url, Tailscale, or plugins.entries.device-pair.config.publicUrl.

    Also in 18 projects that use this

  • molt-botSkill

    discord-e2e

    Test Discord with Convex-leased QA bots: check credentials without a Gateway, exercise owned messages, or prove OpenClaw replies, files, threads, reactions, streaming, and typing. Distinguish API and Gateway evidence from manual user interactions.

    Also in 10 projects that use this

  • molt-botSkill

    openclaw-repair-sweep

    Run scoped OpenClaw issue/PR repair campaigns: coordinate workers, prove root causes, and land or close verified work under the requested authority.

    Also in 10 projects that use this

  • +2 more from molt-bot

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 81 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Add --extract-only when the user explicitly asks for a transcript, not a summary.

    line 33
  2. Very specific:

    Look in received-files/ and received-images/ first — that's where inbound attachments land.

    line 44
  3. Distinctive:

    Keep the whole reply under ~700 chars so it lands in one iMessage chunk.

    line 53
  4. Very specific:

    Default uses Gemini Flash via GEMINI_API_KEY — the harness exports it as EDMUND_GEMINI_KEY, so prepend it to the invocation.

    line 57
  5. Very specific:

    Don't pass --model unless the user asks for a specific one.

    line 58

“Common” is measured: kept by at least 6 in 10 of the 81 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

summarize

Fast inline answer to "what's this link about?" — no page-opening, no "I can't browse" cop-outs.

Triggers

  • "summarize this", "tl;dr", "what's this about"
  • A bare URL followed by any question ("…is this legit?", "worth watching?")
  • "transcribe this YouTube / video"
  • "what does this PDF say"

The three modes

URL / article

GEMINI_API_KEY="$EDMUND_GEMINI_KEY" \
  summarize "https://..." --length short --json

YouTube

GEMINI_API_KEY="$EDMUND_GEMINI_KEY" \
  summarize "https://youtu.be/XXXX" --youtube auto --length short --json

Add --extract-only when the user explicitly asks for a transcript, not a summary. If the transcript is huge, give a summary first and ask which section to expand — don't dump 10k chars into iMessage.

Latency warning. YouTube summarization typically runs 30-120s (transcript fetch + Gemini pass). For videos over 20 minutes it can run 2-4 min. Before calling it, if the video looks long, tell the user upfront: "on it — this'll take a minute." That text will flush only at the end of the turn, so the ideal mitigation is just: start summarize promptly, don't layer extra tool calls on top. Keep --length short (not medium or long) as the default.

Local PDF / file (in the sandbox)

GEMINI_API_KEY="$EDMUND_GEMINI_KEY" \
  summarize "$EDMUND_SANDBOX_PATH/received-files/thing.pdf" --length short --json

Look in received-files/ and received-images/ first — that's where inbound attachments land.

Reply shape (always)

Reformat the model output into:

  • 3 bullets (the substance)
  • 1 line takeaway ("worth it: yes/no because …", "do this: …", "skip: …")

Keep the whole reply under ~700 chars so it lands in one iMessage chunk. If the source is dense, err toward fewer bullets with sharper wording rather than longer ones.

Model + auth

  • Default uses Gemini Flash via GEMINI_API_KEY — the harness exports it as EDMUND_GEMINI_KEY, so prepend it to the invocation.
  • Don't pass --model unless the user asks for a specific one. Flash is fast and cheap for TL;DRs.

Anti-patterns

  • Opening the URL in a browser / web-fetching when summarize can just answer. That's slower and drops detail.
  • Returning the full --json blob. Parse it, reformat to the 3-bullet + takeaway shape.
  • Skipping the takeaway line. The bullets are the what; the takeaway is the so what — that's the part the user actually wanted.
  • Summarizing something the user already has context on (a link they shared and already discussed in-thread). Ask before summarizing redundantly.

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.

Check your setup, free