agentleFS
Sign inStart free
ShmalexMShmalexM/agent-skills/skills/learn-later/SKILL.md

learn-later

Skill0 starsChanged 20 days agoMITView on GitHub

Keep a local queue of topics to learn about later, and turn any of them into a short reading pack in Markdown with SVG diagrams. A pack has a plain-language brief, key terms, a worked example, the current state with dated sources, and self-check questions. Packs are plain files in a folder the user picks once (an Obsidian vault works as is), with optional export to Notion or other doc apps. Use when the user types /learn-later, or says "save this for later", "I want to learn about X later", "remind me to read up on X", "what's on my learn-later list", or "prepare my reading on X".

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    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. Very specific:

    Read the user profile (~/.config/agent-skills/profile.md) if it exists.

    line 10
  2. Distinctive:

    If it says the queue is not set up (exit code 3), ask the user one question: where should reading packs go?

    line 20
  3. Distinctive:

    When the user names a topic, or says "save this" right after asking about something, add it:

    line 36
  4. Very specific:

    When the user says they read a pack, run queue.mjs set "<Topic>" --status read.

    line 63
  5. Distinctive:

    Follow the profile's note rules for the destination (for example, required links or tags).

    line 72
  6. Distinctive:

    Keep packs honest: date everything that can change, cite sources, and keep "Where it stands now" separate from the timeless explanation.

    line 73
  7. Distinctive:

    No personal data in packs beyond what the user gave for the "why".

    line 74

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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

Learn later

The user is curious about something but has no time now. Capture it in seconds; later, prepare a reading pack they can finish in about 10 minutes, with pictures that make it click.

Read the user profile (~/.config/agent-skills/profile.md) if it exists. It may set where packs go, note rules for their notes app (links, tags), and their background.

The helper script is scripts/queue.mjs in this skill's folder. Run it with node.

0. Destination (ask once)

node <this skill's folder>/scripts/queue.mjs where

If it says the queue is not set up (exit code 3), ask the user one question: where should reading packs go? Offer:

  • A notes folder they already use, for example an Obsidian vault folder. Obsidian, Logseq and most Markdown editors read the files as they are.
  • A plain folder, for example ~/learn-later.
  • Notion or another doc app, if this harness has that connector. Packs are still written to a local folder first, then exported (section 4).

Then save the answer, so you never ask again:

node <this skill's folder>/scripts/queue.mjs init --root "<folder>"

The files are always Markdown plus SVG. Do not invent another format.

1. Capture (fast, no research)

When the user names a topic, or says "save this" right after asking about something, add it:

node <this skill's folder>/scripts/queue.mjs add "<Topic>" --why "<one line: what prompted it, from the conversation>"

Pick a short, stable topic name ("Stablecoins", not "that crypto dollar thing"). If you inferred the topic from the conversation, say it in one line so the user can correct it. Then offer: "Prepare the pack now, or leave it queued?" Capture alone takes one tool call; do not start research unless the user wants the pack.

2. Prepare a pack

  1. Scope. One topic, about a 10-minute read: 700 to 1,400 words plus diagrams. If the topic is broad, pick the core idea and list the rest under "Next topics".
  2. Research. For anything factual or current, use primary and reputable sources, and note the date each fact was true. For a bigger topic, hand source-gathering to a cheap or standard worker (the subagent skill) with an output contract of facts, URLs, dates, and checked or inferred. Check every number and date you keep. Never invent figures. Mark simplifications.
  3. Write the pack from references/pack-template.md to the path that queue.mjs path "<Topic>" prints. Write in plain style (the explain-style skill). Lead with the one-line answer and the intuition, then mechanics, then the current state.
  4. Diagrams. Draw 1 to 3 SVG diagrams by following references/diagram-style.md. Usually draw "how it works" (a flow), and add "the landscape" (types or players) or "timeline" when they help. Save them next to the pack and embed them with ![alt text](file.svg). For photos or third-party images, link to them; do not download them.

  5. Check, then mark it ready:

    node <this skill's folder>/scripts/queue.mjs check "<Topic>"
node <this skill's folder>/scripts/queue.mjs set "<Topic>" --status ready

    check makes sure the pack has its sections and sources, and that every embedded SVG exists and parses.

  6. Hand over. Give the absolute path, the read time, and the one-line answer. Mention any open question you could not settle.

3. List and finish

  • queue.mjs list shows the queue (queued, ready, read). --status ready shows what can be read now.
  • When the user says they read a pack, run queue.mjs set "<Topic>" --status read. Offer to add two lines under "What stuck" in the pack; write them only if the user gives them.

4. Export (optional)

If the user chose Notion or a doc app and this harness has that connector, create a page from the pack's Markdown after step 2. Diagrams need care: many doc apps cannot take a local SVG. Attach a PNG of each diagram if the connector supports uploads, or leave a note in the page that points to the local file. Never publish a pack publicly without asking.

Rules

  • The queue and packs belong to the user; never delete entries or packs unless asked.
  • Follow the profile's note rules for the destination (for example, required links or tags).
  • Keep packs honest: date everything that can change, cite sources, and keep "Where it stands now" separate from the timeless explanation.
  • No personal data in packs beyond what the user gave for the "why".

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.

Reviews can't be read right now.

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