agentsclimarketplace

Mind map

Skill loursCreatif/ours-stack/mind-map

Carte mentale interactive — extraction hiérarchique depuis brief/notes/report vers mind-map.json éditable + mind-map.html autonome (pan/zoom, expand/collapse, liens sources). Zero CDN. Presets partagés avec layout-html. Pas de recherche. Déclencheurs : "mind map", "carte mentale", "mind-map", "vue hiérarchique". Commande : /mind-map.From its SKILL.md

Install
npx -y skills add loursCreatif/ours-stack --skill mind-map

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

3 things to look at

  • skips confirmationTells the agent to proceed without asking first, 2 times: "Overwrite without asking unless file modified <1h ago and user didn't say `overwrite`." and 1 more.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 3 commands, including `python3 mind-map/scripts/validate-map.py "<path>/mind-map.json"` and 2 more.

SKILL.md

6.3 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it

Mind Map

Carte mentale cliquable — plus rapide à parcourir qu'un article linéaire, moins immersif qu'un palais 3D.

When to use (vs other skills)

SkillResponsibilityDoes NOT
/layout-htmlArticle linéaire + SVG figuresArbre hiérarchique
/mind-mapThis skill — carte mentale interactiveNouvelle recherche
/deep-researchSynthèse + sources-indexCarte mentale

Optional pipeline (never auto-chained):

/bear-hours → notes / dense-read → /mind-map
/deep-research → /mind-map

Hard rules

  1. No research — never WebSearch / WebFetch. Work only from provided text/files.
  2. Zero external assets — no CDN, no Google Fonts, no external scripts/images/stylesheets, no fetch().
  3. JSON is source of truthmind-map.json editable; HTML regenerated from JSON.
  4. Faithful to source — never invent URLs, claims, or nodes absent from input.
  5. Limits — max 80 nodes, depth 5. Overflow → merge siblings into one detail with bullet note (deterministic, no ask).
  6. Escape output — embed JSON via structured serialization only; never inject raw markdown HTML into template.
  7. Open browseropen <path> on macOS after write.

Step 0: Style — AskUserQuestion

Title: Mind Map — Style

Offer exactly these 4 options (AskUserQuestion max): warm, editorial, minimal, creative. Custom style goes through the automatic "Other" option — treat the free text as a custom preset per references/presets.md § Custom. Smart-skip if one of these presets is named in message.

Default when skipped: warm for all inputs.

Load references/presets.md and token blocks from references/html-template.md.

Step 1: Intake — detect input mode

ModeTriggerSourceOutput
A — Studystudies/<slug>/ or slugbrief.md + optional notes.mdstudies/<slug>/mind-map.{json,html}
B — Researchresearch/<slug>/report.md + optional sources-index.mdresearch/<slug>/mind-map.{json,html}
C — Filepath to .mdthat file<dir>/mind-map.{json,html}
D — Pastedtext in chatmessage bodyoutput/mind-map/<slug>/map.{json,html}
E — Regenerateregenerate / existing JSONmind-map.json onlyregen HTML only

Invocation examples:

/mind-map biomimetisme-locomotion-chantier
/mind-map studies/biomimetisme-locomotion-chantier/brief.md
/mind-map research/robots-assemblage-structurel-chantier
/mind-map regenerate studies/foo/mind-map.json

Guard: no text and no resolvable slug → ask for content or path.

Step 2: Extract hierarchy

Load references/extraction-rules.md and references/node-schema.md.

Parse markdown headings, bullets, source lines. Assign id, type, label, summary, note, href, collapsed, and optional importance.

Use importance only for visual hierarchy: 5 root / central thesis, 4 major branches, 3 supporting concepts, 2 details or examples, 1 sources or minor leaves. Omit it when unsure; the HTML renderer infers a safe size.

Leave top-level layout omitted or "auto" by default. Set "centered" only when the source is clearly wide/exploratory and shallow; set "tree" only when the user explicitly wants the classic hierarchy.

Run validation:

python3 mind-map/scripts/validate-map.py "<path>/mind-map.json"

On validation error → fix JSON before composing HTML.

Step 3: Write JSON then HTML

  1. Write mind-map.json with accurate meta.nodeCount
  2. Compose HTML:
python3 mind-map/scripts/compose-html.py "<path>/mind-map.json"

Or hand-compose from references/html-template.md if script unavailable.

Overwrite without asking unless file modified <1h ago and user didn't say overwrite.

Step 4: Deliver

open "<absolute-path>/mind-map.html"

Tell the user:

  1. Output paths (json + html)
  2. Style preset used
  3. Input mode
  4. Node count by type (concept, detail, example, source); mention summary / note coverage when present
  5. Reminder: edit mind-map.json then /mind-map regenerate <path>

Errors (deterministic)

ConditionAction
Empty / unparseable markdownStop — ask for clearer structure or paste headings
No brief (study mode)Suggest /bear-hours first
>80 nodes after extractionMerge deepest siblings into grouped detail until ≤80
Duplicate idRegenerate ids with slug prefix
source without hrefKeep node; panel shows label only
Invalid JSON (regenerate)Stop — show validate-map.py error
compose-html.py missingFall back to template in references/html-template.md

Escape hatches

User saysAction
regenerateMode E — JSON → HTML only
overwriteReplace without confirmation
warm / editorial / minimal / creativeSmart-skip style question
Described style ("sombre néon", "pastel"…)Smart-skip — custom preset per references/presets.md § Custom (meta.customTokens)
slug onlyTry studies/<slug>/ then research/<slug>/

Self-check before delivery

  • No WebSearch / WebFetch
  • All nodes traceable to input
  • ≤80 nodes, depth ≤5
  • validate-map.py passes
  • HTML offline-ready (no external URLs in assets)
  • JSON + HTML written + browser opened

References

  • references/node-schema.md — JSON contract
  • references/extraction-rules.md — markdown → tree
  • references/html-template.md — manual compose fallback
  • references/presets.md — layout-html token alignment
  • references/design-decisions.md — design decision log
  • scripts/validate-map.py — schema check
  • scripts/compose-html.py — JSON → HTML

What ships with it: 7 files

59.7 KB alongside SKILL.md, 2 of them executable

scripts/

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.