generating-design-references
Generate and critique a design-reference image when no existing visual precedent fits.
Skill by Tom-R-Main
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/tom-r-main-generating-design-references
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Shell scripts
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/generating-design-references
What's inside
- Skill name
generating-design-references- Description
- Generate and critique a design-reference image when no existing visual precedent fits. Use only when an approved image-generation tool or user-configured generator is available; the result is a reasoning scaffold rather than a deliverable or fidelity target.
- License
- MIT
- Instructions
- 37 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (2)
- SKILL.md
- design-reference.sh
SKILL.md (first lines)
# Generating design references Create a concrete visual anchor to think against when a blank-page interface decision cannot be grounded in an existing reference. Installing this skill does not supply image generation, provider authentication, or a cost-free route. ## Ground before generating 1. Search [searching-design-corpus](../searching-design-corpus/SKILL.md) when a corpus is configured and inspect the project's existing UI and user-supplied references. 2. Record why available precedents do not fit. Missing corpus access is a limitation, not evidence that no precedent exists. 3. Generate only when a concrete image will resolve a meaningful design uncertainty. Do not substitute repeated generation for building and exercising the interface. ## Capability preflight Prefer the active host's exposed image tool and follow its runtime skill. Confirm generation is available, the user's scope permits it, and any paid provider invocation is authorized. Do not assume a local subscription transfers to cloud, inspect credentials, install a provider, or claim there is no API cost. When no native tool is exposed, a user may configure `DESIGN_REFERENCE_BIN` to an absolute trusted executable implementing: ```text <generator> --description <text> --output <absolute PNG path> ``` The external generator must use opaque configured runtime credentials, honor the exact output path, and exit nonzero on failure. It is not bundled. The adapter below only invokes that configured executable and verifies the saved file's PNG signature; it supplies no model or authentication. Resolve `SKILL_DIR` from this loaded file. Choose a new output path in the authorized task scratch area: ```bash "$SKILL_DIR/design-reference.sh" "<what to depict, with style cues>" "$RUN_DIR/reference.png" ``` The adapter refuses an existing destination, never scans global image directories, and reports the exact output file. On an uncertain dispatch, inspect the requested file and provider status before retrying. ## View, critique, and verify Open the actual saved image with the host's image-viewing tool. Generation without viewing is not design evidence. Critique hierarchy, spacing, type, components, accessibility, and fit to the product and acceptance criteria. Record what to use and reject. Generated images remain reasoning scaffolds; user-supplied references may be fidelity targets. Never ship a generated mockup as the working interface or treat it as proof of behavior. In a supervised workflow, only the trusted monitor generates, views, critiques, and supplies the image. Writers may inspect the supplied copy but must not invoke the generator. Verify the resulting implementation on its actual surface.
From Tom-R-Main/siftable-skills, MIT.
More from Tom-R-Main/siftable-skills
- 0.592score
paper-design
Tom-R-Main
Create, inspect, edit, import, export, and hand off product or visual designs through Paper and its MCP server.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
grok-guide
Tom-R-Main
Use Grok 4.5 as an explicit strategic control plane above Codex for plan, replan, and completion-audit lifecycles.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
plan
siftable
Produce a visual, grounded plan for an objective.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
filesystem-lifecycle
Tom-R-Main
Audit, classify, reclaim, and prevent regrowth of local filesystem cruft and agent-workspace storage.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
footwork-tools
Tom-R-Main
Equip an agent with observation and action tools for browser UI, native Mac apps, terminals/services, and saved artifacts.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
pgvector-postgres
Tom-R-Main
Design, review, migrate, measure, or debug PostgreSQL vector retrieval with pgvector.
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
go
Tom-R-Main
Implement, review, debug, test, migrate, and optimize Go.
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
c-systems-programming
Tom-R-Main
Implement, debug, and review C systems software.
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
css
Tom-R-Main
Implement, debug, and review CSS and Tailwind styling.
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
opengl-c-graphics
Tom-R-Main
Build, debug, review, and optimize core-profile OpenGL in C.
SkillAFreshness: A★ 1Claude Code +9 - 0.587score
mermaid
siftable
Create Mermaid diagrams and render them through an available host renderer.
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
cloud-environment-runtime
Tom-R-Main
Establish the selected cloud task's execution, network, and credential context before using cloud tools.
SkillAFreshness: A★ 1Claude Code +9
Signals
- Stars
- 1
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 12, 2026 (16 days ago)
- Last release
- Aug 14, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/Tom-R-Main/siftable-skills
- Website
- siftable.io
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from Tom-R-Main
- 0.592score
footwork-tools
Tom-R-Main
Equip an agent with observation and action tools for browser UI, native Mac apps, terminals/services, and saved artifacts.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
pgvector-postgres
Tom-R-Main
Design, review, migrate, measure, or debug PostgreSQL vector retrieval with pgvector.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
grok-guide
Tom-R-Main
Use Grok 4.5 as an explicit strategic control plane above Codex for plan, replan, and completion-audit lifecycles.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
filesystem-lifecycle
Tom-R-Main
Audit, classify, reclaim, and prevent regrowth of local filesystem cruft and agent-workspace storage.
SkillAFreshness: A★ 1Claude Code +9 - 0.592score
paper-design
Tom-R-Main
Create, inspect, edit, import, export, and hand off product or visual designs through Paper and its MCP server.
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
css
Tom-R-Main
Implement, debug, and review CSS and Tailwind styling.
SkillAFreshness: A★ 1Claude Code +9
Cite this page
generating-design-references skill. Agent Mart. https://agentmart.dev/skills/tom-r-main-generating-design-references (last updated 2026-10-24)