autonomous-workshop: Skill for Claude Code

.claude/skills/brainstorm-concepts/SKILL.md

brainstorm-concepts is a skill for Claude Code from autonomous-ai/autonomous-workshop. It costs 82 tokens per session (2,134 once invoked), scanned A, original, Apache-2.0.

A guided process for exploring toy or product ideas through batches of generated images. A person compares the images, gives feedback, and chooses a concept.

In plain words
What is it for?
Use it to explore shapes and styles for a new toy or product, change the look of an existing product, or redraw concepts in a different style.
Why use it?
It makes early visual exploration concrete instead of relying only on written descriptions. Each round uses the previous feedback to shape the next set of ideas.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: reads .claude/ paths.

This is autonomous-ai/autonomous-workshop's own configuration. It tells Claude Code how to work on autonomous-workshop itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything autonomous-workshop configures →

Reuse

Borrowing it

Nothing to install: this file belongs to autonomous-ai/autonomous-workshop. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/autonomous-ai/autonomous-workshop/main/.claude/skills/brainstorm-concepts/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/autonomous-ai/autonomous-workshop

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for brainstorm-concepts

README.md
[![agentmods](https://agentmods.dev/badge/skills/autonomous-ai/autonomous-workshop/brainstorm-concepts/github.svg)](https://agentmods.dev/skills/autonomous-ai/autonomous-workshop/brainstorm-concepts)
Your own site
<a href="https://agentmods.dev/skills/autonomous-ai/autonomous-workshop/brainstorm-concepts"><img src="https://agentmods.dev/badge/skills/autonomous-ai/autonomous-workshop/brainstorm-concepts/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for brainstorm-concepts

Your own site · 80×15
<a href="https://agentmods.dev/skills/autonomous-ai/autonomous-workshop/brainstorm-concepts"><img src="https://agentmods.dev/badge/skills/autonomous-ai/autonomous-workshop/brainstorm-concepts.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 82 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,134 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00082 $0.02134
Opus 5.5 $0.00033 $0.00854
Sonnet 5.5 $0.00016 $0.00427
Haiku 4.5 $0.00008 $0.00213

Measured 7d ago against content hash aac04edd62e0, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

brainstorm-concepts scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/embed_page.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/brainstorm-concepts/SKILL.md · 156 lines

How it starts

The opening of the file, as written. The whole thing — 156 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Brainstorm concepts

A Concept Round is a batch of Concept Images (both terms are in CONTEXT.md) for the human to look at and react to. Their reaction shapes the next round. The human picks by picture, so pictures are the only thing you ask them to judge. The skill ends when they mark a concept chosen. It decides no dimension, writes no contract and hands nothing off.

Setup, once per subject

  1. Facts are your job. Read what exists about the subject yourself: its product page, its folder under toys-spec/ and auto-memory. Ask the human only for decisions, such as which product they mean.
  2. Work in toys-spec/<subject>/. Write the constraints every image must honour at the top of NOTES.md. They come from the subject's job (for example, a phone taps it, so it needs a solid tap area of at least 28 mm), its size and its weight. Under them, start an empty Rejections log.

Done when: NOTES.md states the subject, its job and its constraints, and has a Rejections log.

A Concept Round

  1. Plan. Make 6–8 concepts unless the human names a count or a split ("6 tech, 6 not"). Every concept is a distinct idea. Give the round a letter prefix not used before (K1, R1, C1…) so every concept keeps a unique name. Write one shared prompt block and one line per concept:
    • The shared block carries the constraints and a studio shot: one standalone object fully inside the frame with margin, a plain surface, soft daylight, a three-quarter view from above, and no text or other objects. It also carries an explicit exclusion for everything the human has rejected, from the Rejections log in NOTES.md.
    • The shared block also limits plastic to the stocked filaments: list the colours in src/workshop/wish/catalogues/filaments.json by name and say that every plastic surface is one flat colour from that list (ADR 0092). Paper areas - pictures, book covers and spines, wallpaper, skies, notices - may use any colour, because they are printed.
    • The concept line names the idea, its form, its colours and the detail that makes it this concept. Each plastic colour it names is one from the list.
  2. Redraw from images when the human asks for one concept in another's style, or sends a picture. Attach images with --ref, the concept to keep first and the style to copy second, and say so in the prompt ("the first image is the concept to keep…"). Save every image the human pasted into concepts/round<N>/inputs/. Only the human's images and your own Concept Images are ever attached; an image found on the web is looked at only.
  3. Generate every concept in parallel, each with its own prompt file:
    uv run --no-project --with pillow python \
      .claude/skills/brainstorm-trend/scripts/generate_image.py \
      --prompt-file <prompt.txt> --out toys-spec/<subject>/concepts/round<N>/<id>_<slug>.png \
      --env <.env> [--ref <concept> --ref <style>]
    
    The .env holds OPENROUTER_API_KEY and OPENROUTER_IMAGE_MODEL. A worktree has none, so point --env at the main checkout's .env. Never print the key. Read the printed path, because the suffix follows the decoded format.
  4. Check every image against its concept line: put them all on one contact sheet and look. If an image misses the point of its concept (the wrong count or layout, a hollow tap area, a banned element), regenerate it once. Then take the image as it stands and write its flaw into its caption and NOTES.md.
  5. Page. toys-spec/<subject>/CONCEPTS.html is one page in Vietnamese, with the newest round on top and earlier rounds below it. Each concept is one card: its picture, its id and name, an approximate size, one line on the idea, any flaw, and a Đã chọn badge on chosen concepts. Publish it with the Artifact tool, and republish the same file path so the link stays the same. If the artifact was deleted, ask before publishing a new link. If the Artifact tool is off, skip the publish; the standalone copy below still happens.
  6. Standalone copy, every time the page changes, unasked. The human opens pages from the main checkout, where the artifact's relative image paths may not resolve. Write a copy with every image inlined next to the page:
    uv run --no-project --with pillow python \
      .claude/skills/brainstorm-concepts/scripts/embed_page.py \
      toys-spec/<subject>/CONCEPTS.html toys-spec/<subject>/CONCEPTS-standalone.html
    
    Do not commit or push it; run tools/sync-to-main.sh so it reaches the main checkout now. In the report, give its absolute path in the main checkout, never a relative path.
  7. Notes. Append a round section to NOTES.md:
    • the human's ask, in their words;
    • each concept id with its idea;
    • the source of every image (text only, or which --ref inputs);
    • every flaw.
  8. Do not commit the round's sources. New files under toys-spec/ are gitignored; tools/sync-to-main.sh (the project's Stop hook) copies them into the main checkout. Only files Git already tracks are committed and pushed.
  9. Report in the human's language, briefly: a table of id, name and idea, then the flaws, two or three picks with a one-line reason each, and one question.

Read the full file on GitHub · 156 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 7d ago Changed · +46 lines aac04edd62e0
  2. 22d ago Changed · +2 lines dbb154f667f8
  3. 27d ago First seen · 108 lines · 82 tokens per session scan A 6a437bc3ddb1

Subscribe to this mod's changes

brainstorm-concepts is a skill published in the GitHub repository autonomous-ai/autonomous-workshop (24 stars, last pushed yesterday), licensed Apache-2.0. It adds 82 tokens to every session and 2,134 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-10.

Related

brainstorm-concepts compared with skills from other repositories

brainstorm-concepts next to 6 skills from other repositories
ModTypeStarsLast pushLicenceScanTokensWorks with
brainstorm-conceptsautonomous-ai/autonomous-workshopSkill24yesterdayApache-2.0A82Claude Code
cocoindex-diagramscocoindex-io/cocoindexSkill12kyesterdayApache-2.0A124Claude Code, Codex
cli-commandComposioHQ/composioSkill31kyesterdayMITA76Claude Code, Codex
dashboard-designmckinsey/vizroSkill3.8kyesterdayApache-2.0A71Claude Code
dstack-prototypingdstackai/dstackSkill2.3kyesterdayMPL-2.0A80Claude Code, Codex
designing-vizro-layoutsmckinsey/vizroSkill3.8kyesterdayApache-2.0A58Claude Code
diagram-designvstorm-co/pydantic-deepagentsSkill1.1k7d agoMITA16Claude Code, Codex