Back to Skills

render-icon-pipeline

pjt222
Updated 2 days ago
5 views
17
2
17
View on GitHub
Metadesigndata

About

This skill runs a visualization pipeline to render icons from existing glyphs for skills, agents, and teams. It handles palette generation, data building, manifest creation, and icon rendering. Developers must always use the `build.sh` script as the entry point, never calling Rscript directly.

Quick Install

Claude Code

Recommended
Primary
npx skills add pjt222/agent-almanac -a claude-code
Plugin CommandAlternative
/plugin add https://github.com/pjt222/agent-almanac
Git CloneAlternative
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/render-icon-pipeline

Copy and paste this command in Claude Code to install this skill

Documentation

Render Icon Pipeline

Run viz pipeline end-to-end → render icons from existing glyphs. Palette gen, data build, manifest create, icon render for skills, agents, teams.

Canonical entry: bash viz/build.sh [flags] from project root, or bash build.sh [flags] from viz/. Script handles platform detection (WSL, Docker, native), R binary select, step ordering. Never call Rscript direct for build scripts — that path only for MCP server config.

Use When

  • After create/modify glyph fns
  • After add new skills, agents, teams to registries
  • Icons need re-render for new/updated palettes
  • Full pipeline rebuild (after infra changes)
  • Setting up viz env first time

In

  • Optional: Entity type — skill, agent, team, all (default all)
  • Optional: Palette — specific name or all (default all)
  • Optional: Domain filter — specific domain for skill icons (e.g. git, design)
  • Optional: Render mode — full, incremental, dry-run (default incremental)

Do

Step 1: Verify Prereqs

Ensure env ready for rendering.

  1. Confirm viz/build.sh exists:
    ls -la viz/build.sh
    
  2. Verify Node.js available:
    node --version
    
  3. Check viz/config.yml exists (platform-specific R path profiles):
    ls viz/config.yml
    

build.sh handles R binary resolution auto — no need verify R paths manually. WSL → /usr/local/bin/Rscript (WSL-native R), Docker → container R, native Linux/macOS → Rscript from PATH.

build.sh, Node.js, config.yml present.

If err: config.yml missing → pipeline falls back to system defaults. Node.js missing → install via nvm.

Step 2: Run Pipeline

build.sh exec 5 steps in order:

  1. Generate palette colors (R) → palette-colors.json + colors-generated.js
  2. Build data (Node) → skills.json
  3. Build manifests (Node) → icon-manifest.json, agent-icon-manifest.json, team-icon-manifest.json
  4. Render icons (R) → icons/ + icons-hd/ WebP files
  5. Generate terminal glyphs (Node) → cli/lib/glyph-data.json

Full pipeline (all types, all palettes, std + HD):

bash viz/build.sh

Incremental (skip icons existing on disk):

bash viz/build.sh --skip-existing

Single domain (skills only):

bash viz/build.sh --only design

Single entity type:

bash viz/build.sh --type skill
bash viz/build.sh --type agent
bash viz/build.sh --type team

Dry run (preview no rendering):

bash viz/build.sh --dry-run

Std size only (skip HD):

bash viz/build.sh --no-hd

All flags after build.sh passed through to build-all-icons.R.

→ Icons rendered to viz/public/icons/<palette>/ + viz/public/icons-hd/<palette>/.

If err:

  • renv hang on NTFS: viz .Rprofile bypasses renv/activate.R + sets .libPaths() direct. Ensure run from viz/ (build.sh does auto via cd "$(dirname "$0")")
  • Missing R pkgs: Run Rscript -e "install.packages(c('ggplot2', 'ggforce', 'ggfx', 'ragg', 'magick', 'future', 'furrr', 'digest'))" from R env build.sh selects
  • No glyph mapped: Entity needs glyph fn — use create-glyph before rendering

Step 3: Verify Output

Confirm render completed.

  1. Check file counts match:
    find viz/public/icons/cyberpunk -name "*.webp" | wc -l
    find viz/public/icons-hd/cyberpunk -name "*.webp" | wc -l
    
  2. Check reasonable file sizes (2-80 KB per icon)
  3. Run audit-icon-pipeline for comprehensive check

→ File counts match manifest entry counts. Sizes in expected range.

If err: counts don't match → some glyphs errored during render. Check build log for [ERROR] lines.

CLI Flag Reference

All flags passed through build.shbuild-all-icons.R:

FlagDefaultDescription
--type <types>allComma-separated: skill, agent, team
--palette <name>allSingle palette or all (9 palettes)
--only <filter>noneDomain (skills) or entity ID (agents/teams)
--skip-existingoffSkip icons with existing WebP files
--dry-runoffList what would be generated
--size <n>512Output dimension in pixels
--glow-sigma <n>4Glow blur radius
--workers <n>autoParallel workers (detectCores()-1)
--no-cacheoffIgnore content-hash cache
--hdonEnable HD variants (1024px)
--no-hdoffSkip HD variants
--strictoffExit on first sub-script failure

What build.sh Does Internally

For ref only — do NOT run these manually:

cd viz/
# 1. Platform detection: sets R_CONFIG_ACTIVE (wsl, docker, or unset)
# 2. R binary selection: WSL → /usr/local/bin/Rscript, Docker → same, native → Rscript
# 3. $RSCRIPT generate-palette-colors.R
# 4. node build-data.js
# 5. node build-icon-manifest.js --type all
# 6. $RSCRIPT build-all-icons.R "$@"  (flags passed through)
# 7. node build-terminal-glyphs.js

Docker Alternative

Pipeline can also run in Docker:

cd viz
docker compose up --build

Runs full pipeline in isolated Linux env + serves on port 8080.

Check

  • Ran bash viz/build.sh (not bare Rscript)
  • Palette colors generated (JSON + JS)
  • Data files built from registries
  • Manifests generated from data
  • Icons rendered for target types + palettes
  • File counts match
  • Sizes in expected range (2-80 KB)

Traps

  • Calling Rscript direct: Never Rscript build-icons.R or Rscript generate-palette-colors.R manually. Always bash build.sh [flags]. Direct calls bypass platform detection + may use wrong R binary (Windows R via ~/bin/Rscript wrapper instead of WSL-native R at /usr/local/bin/Rscript). Note: Windows R path in CLAUDE.md + guides for MCP server config only, not build scripts.
  • Wrong cwd: build.sh CDs to own dir auto (cd "$(dirname "$0")"), so call from anywhere: bash viz/build.sh from project root works.
  • Stale manifests: build.sh runs Steps 1-5 in order, manifests always regen before render. Only need manifests no render → node viz/build-data.js && node viz/build-icon-manifest.js (Node steps no need R).
  • renv not activated: .Rprofile workaround needs running from viz/build.sh handles. Using --vanilla or running R from another dir skips it.
  • Parallel on Windows: Windows no support fork-based parallelism — pipeline auto-selects multisession via config.yml.

GitHub Repository

pjt222/agent-almanac
Path: i18n/caveman-ultra/skills/render-icon-pipeline
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

Related Skills

content-collections

Meta

This skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.

View skill

polymarket

Meta

This skill enables developers to build applications with the Polymarket prediction markets platform, including API integration for trading and market data. It also provides real-time data streaming via WebSocket to monitor live trades and market activity. Use it for implementing trading strategies or creating tools that process live market updates.

View skill

creating-opencode-plugins

Meta

This skill helps developers create OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It provides the plugin structure, event API specifications, and implementation patterns for JavaScript/TypeScript modules. Use it when you need to intercept, monitor, or extend the OpenCode AI assistant's lifecycle with custom event-driven logic.

View skill

sglang

Meta

SGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.

View skill