agentleFS
Come-In-LabsCome-In-Labs/agent-skills/skills/visualize/SKILL.md

visualize

Skill2 starsChanged 9 days agoMITView on GitHub

Use when someone wants a design, feature, flow or data model shown as real diagrams instead of prose - "visualisiere das", "zeig mir die Abläufe als Diagramme", "draw the flows", "show the schema", "before and after", "what changes in the data model", "what can be dropped or shrunk", "side by side today vs new", "Datenschema aufzeigen" - for a spec, a plan, a redesign or an existing feature, usually as a page to share. NOT a single quick chart of numbers (use dataviz), NOT a UI mockup (use a design tool), NOT a review of the design itself (use cil:roast or cil:cross-examine).

46 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use exp-pilot-eval, exp-pilot-run 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.
  • skyllwtSkill

    exp-pilot-eval

    Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.

    Also in 4 projects that use this

  • skyllwtSkill

    exp-pilot-run

    Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.

    Also in 4 projects that use this

  • LambenthanSkill

    daily-arxiv

    每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新

    Also in 4 projects that use this

  • LambenthanSkill

    exp-eval

    实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

    Also in 4 projects that use this

  • phanghonghaoSkill

    poster

    HTML/CSS A4 海报生成。从沙龙/活动 Markdown 文件生成科技风 A4 海报 PDF/PNG。Use when user mentions: poster, 海报, 海报生成, make poster, generate poster, 活动海报, 沙龙海报.

    Also in 4 projects that use this

  • +1 more from skyllwt
  • +6 more from Lambenthan

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 44 edited versions we read keep these.

  • 5

    Distinctive

    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
  1. Distinctive:

    Before drawing anything, write down what the figures will claim and check each claim in the source:

    line 18
  2. Distinctive:

    every table and column you will show, read from the schema files

    line 21
  3. Distinctive:

    Keep the file:line for each.

    line 26
  4. Distinctive:

    Lay out on fixed columns and rows: Write the coordinates down before the markup.

    line 56
  5. Very specific:

    Run a script over the page: balanced <text> tags, no duplicate ids, none of the project's banned characters.

    line 64
  6. Distinctive:

    Look at the page once in light and dark mode at phone and desktop width, if the session offers a preview.

    line 65

“Common” is measured: kept by at least 6 in 10 of the 44 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

cil:visualize - flows and schema as diagrams that say something

What this is

A diagram earns its place when a cold reader sees a mechanism they would otherwise assemble from prose: who decides, what gets stored, where two records exist for one thing, what disappears. A box per noun with arrows between them says nothing the spec did not.

The page shows today and after side by side, wherever something changes. Every statement about today is read in the code before it is drawn. Proposals are labelled as proposals.

1. Collect the facts first

Before drawing anything, write down what the figures will claim and check each claim in the source:

  • the stored values behind every state (enum or status column values, not invented labels)
  • every table and column you will show, read from the schema files
  • who triggers each transition (a person, the system, a webhook, a cron job)
  • counts you will print ("20 files read this table"), measured with a grep you can repeat
  • columns or tables that nothing reads or writes; a grep for readers and writers proves it

Keep the file:line for each. A wrong fact in a confident diagram does more damage than a missing diagram.

2. Pick the figures

Draw only what the decision turns on. Most designs need four to seven figures from this set:

FigureShowsMust carry
Ways inInputs to the stores they land inEdge labels saying what happens automatically, and who decides
LifecycleStates of the main recordThe stored value under each state; edges labelled by actor or trigger; the transition nobody approves, marked
Who creates whatOne flow, step by stepPer step: an actor pill, the action, and the record that comes into being. Duplicates become visible here
ComputationHow a derived result is builtStages in order, sources feeding in, what is stored and what is only computed
Schema todayTables that move, columns that matterWhat goes away, marked × in the warning colour, with the phase it goes in
Schema afterThe same tables after the changeNew columns +, changed ~, each tagged with its phase; a "goes away" box
Shrink tableEverything that can be dropped or made smallerToday (verified), after, when; what can go now without waiting for a phase

Pair today and after for every figure where both exist. A "new" figure without its "today" hides what the change costs. For a feature that is already built, "after" is the trimmed version the shrink table proposes.

Pick figures from the reader's question, not from the code's structure. The reader is usually the person who decides what the product does. A component tree, a call graph of server functions or a request sequence through every helper describes the code, and belongs on the page only when the question is about the code's structure.

The page is, in this order: a three-line summary with the legend; the figures, today next to after; the schema pair; the shrink table; the phases with their schema change, when the work ships in steps; one line naming the spec, plan and other pages it draws from. Test coverage maps, file indexes and decision lists are not part of it.

3. Draw by hand, as inline SVG on a page

Publish an Artifact page. REQUIRED SUB-SKILL: load artifact-diagramming for the SVG mechanics and the page contract, then:

  • Theme through page CSS classes on the SVG elements (.n, .n-new, .n-warn, .n-gone, .e, .el), each reading colour tokens defined on :root with dark-mode overrides. No literal colours inside the SVG.
  • One state language across all figures: unchanged, new or changed, a problem today, goes away (dashed in the warning colour), computed and not stored (dashed neutral). Show it once in a legend at the top.
  • Give every marker id a per-figure prefix (f2a-arrow). Ids are page-wide, and a duplicate silently breaks arrowheads in the other figures.
  • Put wide figures in an overflow-x: auto wrapper with a min-width on the SVG, so the page itself never scrolls sideways.
  • Lay out on fixed columns and rows. Write the coordinates down before the markup. Schema boxes get a header band plus 18 px per row.
  • Draw FK lines only between neighbouring columns, routed orthogonally through the gap. Farther relations go into the row text as client_id → clients. Twenty crossing lines read as noise.
  • Budget the text. At these sizes, mono 10.5 px is about 6.3 px per character, body 11 px about 5.5, and bold 12.5 px about 7. Shorten a label rather than let it run past its box.
  • Keep edge labels off the line. For a curve, estimate the point at the label's height and set the label beside it, not on it.
  • Write a caption under each figure that states its claim and what the drawing leaves out, such as an edge drawn only in the caption or an exception.

4. Check before publishing

  • Run a script over the page: balanced <text> tags, no duplicate ids, none of the project's banned characters.
  • Look at the page once in light and dark mode at phone and desktop width, if the session offers a preview.
  • Re-read every number and name on the page against your step 1 notes.

Common mistakes

MistakeInstead
Labels that restate box names ("Booking → Bookings")Label the edge with the action or rule: paid: confirmed, nobody reviewed it
Only the new design drawnToday and after side by side, with the same layout
State names invented for the pictureThe stored value under each state
Full column dumps in the schemaOnly the columns that move or that the argument needs
A schema line to every FKLines for neighbours, → table text for the rest
Mermaid or a chart library for flowsHand SVG; theming and label placement stay under your control
"Can be dropped" without proofA grep showing no reader and no writer, cited in the shrink table
Proposed names shown as if they existSay in the caption which names are proposals

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.