distill
Skill0 starsChanged 3 months agoMITView on GitHub
Turns raw material — a meeting transcript, a PDF, a pasted pile of notes, a call recording summary, a document someone sent over — into routed, linked knowledge instead of one giant dumped file. Fires on requests like "distill this transcript", "process these meeting notes", "I just got off a call, here's the dump", "turn this into knowledge", "file this away properly", "my knowledge base is just a pile of unprocessed transcripts", "I keep pasting raw notes and nothing happens to them", "where does this go", or "ingest this document". Also fires when the user pastes a large block of raw text (a transcript, an email thread, a call recap) with no further instruction — that paste is the raw material, not a request to summarize it back to them.
76 other projects use a version of this.
Adopted by
Most recently taken up by
- DeepExtrema/Tesseracttheir own version, first seen Nov 16, 2026
- persiyanov/skillstheir own version, first seen Nov 15, 2026
- komluk/scaffoldingtheir own version, first seen Nov 15, 2026
- Integralist/agent-skillstheir own version, first seen Nov 14, 2026
- theexperiencecompany/gaiatheir own version, first seen Nov 13, 2026
- venhdev/skillstheir own version, first seen Nov 8, 2026
Used with
Projects with it also use layout, shape 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.
owshq-mecSkilllayout
Improve layout, spacing, and visual rhythm. Fixes monotonous grids, inconsistent spacing, and weak visual hierarchy. Use when the user mentions layout feeling off, spacing issues, visual hierarchy, crowded UI, alignment problems, or wanting better composition.
Also in 8 projects that use this
HAR5HA-7663Skillshape
Plan the UX and UI for a feature before writing code. Runs a structured discovery interview, then produces a design brief that guides implementation. Use during the planning phase to establish design direction, constraints, and strategy before any code is written.
Also in 11 projects that use this
ReScienceLabSkillarchive
Archive session learnings, debugging solutions, and deployment logs to .archive/yyyy-mm-dd/ as indexed markdown with searchable tags. Use when completing a significant task, resolving a tricky bug, deploying, or when the user says \"archive this\". Maintains .archive/MEMORY.md index for cross-session knowledge reuse.
Also in 3 projects that use this
klauswgSkillckm:design-system
Token architecture, component specifications, and slide generation. Three-layer tokens (primitive→semantic→component), CSS variables, spacing/typography scales, component specs, strategic slide creation. Use for design tokens, systematic design, brand-compliant presentations.
Also in 3 projects that use this
termdockSkillrecall
Retrieves cross-session memories about past decisions, lessons, and patterns. Use when user asks about prior work, past decisions, or mentions a feature. Triggers include check memory, what did we decide, how did we solve.
Also in 6 projects that use this
iovisorSkillevolve
Evolve uBPF through a gated spec-first workflow that propagates approved changes through requirements, design, validation, implementation, and RFC alignment artifacts.
Also in 3 projects that use this
Other versions76 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 76
- juxt/allium★ 506same name
- educlopez/ui-craft★ 441same name
- Michaelliv/napkin★ 428same name
- nexmoe/DayBee★ 424same name
- nexmoe/eve★ 424same name
- Southclaws/storyden★ 351same name
- 10xChengTu/Mux0★ 337same name
- theexperiencecompany/gaia★ 327same name
- AkaraChen/aghub★ 271same name
- zacharyfmarion/openscad-studio★ 252same name
66 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 3 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 72 edited versions we read keep these.
3
Distinctive
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
- Distinctive: line 17
Read the raw input in full: Don't skim for a summary — you're about to split it into pieces, and you can't split what you haven't read.
- Distinctive: line 28
Use today's date, not the date of the source material, unless the operator says otherwise.
- Distinctive: line 41
If a piece of material doesn't fit any of the four, it's probably not worth keeping — raw material is not automatically knowledge.
- Very specific: line 44
Check for contradictions before you write: Before adding or editing a node, search knowledge/external/ and knowledge/playbook/ for existing …
- Very specific: line 47
Do not silently overwrite it.
“Common” is measured: kept by at least 6 in 10 of the 72 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.
Files
Distill
The failure this prevents: raw material gets pasted whole into a knowledge file. Nothing gets
decided, nothing gets routed, nothing gets linked — a transcript just becomes a longer file.
Six months later the agent reads four thousand words to find one fact, most sessions never
read it at all, and the one decision buried on page three of the transcript never made it to
decisions/log.md. That's bloat plus a decision that, for every practical purpose, was
never made.
Procedure
- Read the raw input in full. Don't skim for a summary — you're about to split it into pieces, and you can't split what you haven't read. Never copy it verbatim into a knowledge file. If you cannot tell what kind of material this is (transcript, document, notes), ask before routing.
Route into four destinations, not one. Most of the value in raw material is destroyed by filing it in a single place — a transcript is never one kind of thing. Go through the material looking for each of these separately, because a single paragraph can contain more than one:
- Decisions — something was decided, chosen, ruled out, or committed to. Append to
decisions/log.mdin the format[YYYY-MM-DD] DECISION: … | REASONING: … | CONTEXT: …. Use today's date, not the date of the source material, unless the operator says otherwise. - Durable facts about the outside world — a client, a prospect, a competitor, a market
condition, something true about someone or something outside the operator's own company.
Goes to
knowledge/external/as one or more knowledge nodes (see step 4). - The operator's own reusable method — a principle, a play, a way of doing the work that
will apply again next time, not just this once. Goes to
knowledge/playbook/as one or more knowledge nodes (see step 4). - Action items — something someone needs to go do. These do not go in a knowledge file
at all. Check the precedence table in
CLAUDE.mdfor "Task and project status" — that names the task tool this operator actually uses (ClickUp, Linear, Notion, or similar). Send action items there. If no task tool is configured, say so and ask where they should go rather than inventing a file for them.
If a piece of material doesn't fit any of the four, it's probably not worth keeping — raw material is not automatically knowledge. Say what you're leaving out and why.
- Decisions — something was decided, chosen, ruled out, or committed to. Append to
- Check for contradictions before you write. Before adding or editing a node, search
knowledge/external/andknowledge/playbook/for existing claims on the same subject. If the new material contradicts one:- Do not silently overwrite it. That's
clash, and it destroys whichever version was right without leaving a trace that there was ever a question. - Surface both versions to the operator and ask which one wins.
- Record the resolution in
decisions/log.mdin the same format as step 2, regardless of which side won — the fact that a contradiction was found and resolved is itself worth keeping.
- Do not silently overwrite it. That's
Write each knowledge node as one claim, not one topic. A node is a single fact or principle, with its source and its date, linked to its neighbors. Don't write a node titled after the meeting ("2026-08-27 call notes") — write one node per claim that survives the meeting. Shape:
# <the claim, stated plainly> <Two to five sentences. What this means and why it matters. Nothing the reader has to dig for.> **Source:** <where this came from — call, document, person> **Date:** <YYYY-MM-DD> Related: [[other-node-slug]], [[another-node-slug]]A forty-minute call might produce zero nodes, one node, or five — it depends on how many distinct claims survive. It almost never produces one node per transcript.
- Update the wiki's
index.mdin the same pass — not as a follow-up. Every new file inknowledge/external/orknowledge/playbook/gets an entry in that folder'sindex.mdbefore you consider the work done. An index that lags the nodes it's supposed to list isconfusion: the next session reads the index, doesn't see the node, and reports that the knowledge doesn't exist — even though the file is sitting right there unlinked. - Say where the raw file should live, and leave it there. This kit does not impose a
folder for raw source material — that's a deliberate gap, not an oversight. Tell the
operator where you'd put it (next to the project it belongs to, or wherever they keep source
material) and let them place it. Don't invent a new top-level folder to hold it — an
undocumented folder is exactly the kind of drift
/os-auditexists to catch. - Report what you did, not what you read. Close with a short list: which decisions got logged, which nodes got written or updated (with their paths), which contradictions got surfaced, and where the action items went. If nothing in the material warranted any of the four routes, say that plainly instead of manufacturing a node to justify the pass.
When NOT to run this
/route-fix is for when the agent already had knowledge somewhere and failed to find it —
a routing, index, or precedence problem. This skill is for when the knowledge doesn't exist
in routed form yet and needs to be added. If the operator says "the agent should have known
this" about something already on disk, that's /route-fix; if they're handing you something
new to process, that's this skill.
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.

