skills4sh
v0.4.14
Published
Agent skills for Claude Code, Cursor, and VS Code (Copilot).
Readme
skills4sh
Agent skills for Claude Code, Cursor, and VS Code (Copilot).
Skills
| Skill | Description | Version | |-------|-------------|---------| | agent-memory | Cross-interface persistent memory system for any project | 2.7.6 | | code-to-figma | Generate a CI-anchored code → Figma token export pipeline (walker + Gist) that the tokens-sync-to-figma plugin consumes | 0.1.5 | | discord-harvest | Extract and download images, links, and files from Discord conversations | 1.7.7 | | eleventy-nunjucks | 11ty v3 + Nunjucks conventions, syntax cheat sheet, autoescape rules, stability + security checklists | 0.1.7 | | figma-to-code | Repo-first Figma MCP workflow for UI implementation, tokens, design-system rules, and Code Connect | 0.1.6 | | localhost-screenshots | Localhost screenshot capture and visual regression testing | 3.3.6 | | skill-architect | Portable skill authoring/review rubric that bridges Skill Development, writing-skills, Anthropic/OpenAI skill-creator patterns, vendor adapters, and eval planning | 0.1.2 |
Stability note.
code-to-figma,eleventy-nunjucks,figma-to-code, andskill-architectare pre-1.0 (v0.x) — their trigger phrasing and prompt content may change in incompatible ways between minor releases. The other three skills (agent-memory, discord-harvest, localhost-screenshots) follow semver: breaking changes only on major bumps.
Install
These skills are plain Agent Skills: each skill is a folder with a SKILL.md entrypoint plus optional references/assets. Use any installer or agent runtime that can place those folders where your agent reads skills.
Recommended install path:
npx skills add t4sh/skills4sh # install all skills
npx skills add t4sh/skills4sh --skill agent-memory # only agent-memory
npx skills add t4sh/skills4sh --skill code-to-figma # only code-to-figma
npx skills add t4sh/skills4sh --skill discord-harvest # only discord-harvest
npx skills add t4sh/skills4sh --skill eleventy-nunjucks # only eleventy-nunjucks
npx skills add t4sh/skills4sh --skill figma-to-code # only figma-to-code
npx skills add t4sh/skills4sh --skill localhost-screenshots # only localhost-screenshots
npx skills add t4sh/skills4sh --skill skill-architect # only skill-architectRe-running the same command is safe and can be used to sync the installed skill copy with this repository.
For a noninteractive global/user-level install, pass the target agent explicitly:
npx skills add t4sh/skills4sh -g -a codex -yReplace codex with your agent/runtime name when supported, such as amp, cline, kimi-code-cli, opencode, warp, or zed. The exact reload/restart behavior depends on the agent.
Uninstall
npx skills remove agent-memory # remove agent-memory
npx skills remove code-to-figma # remove code-to-figma
npx skills remove discord-harvest # remove discord-harvest
npx skills remove eleventy-nunjucks # remove eleventy-nunjucks
npx skills remove figma-to-code # remove figma-to-code
npx skills remove localhost-screenshots # remove localhost-screenshots
npx skills remove skill-architect # remove skill-architectAdd -g to remove from global/user scope. skills remove --all is intentionally not listed: it can remove skills beyond this repository, depending on the installer and destination.
The published skills4sh package is a supporting installer for environments that cannot use the generic skills CLI, need explicit destination control, need no-git installation, or need to pin a specific ref.
npx skills4sh --all # install all skills
npx skills4sh --skill agent-memory # only agent-memory
npx skills4sh --skill code-to-figma # only code-to-figma
npx skills4sh --skill discord-harvest # only discord-harvest
npx skills4sh --skill eleventy-nunjucks # only eleventy-nunjucks
npx skills4sh --skill figma-to-code # only figma-to-code
npx skills4sh --skill localhost-screenshots # only localhost-screenshots
npx skills4sh --skill skill-architect # only skill-architectUninstall (v0.4.0+):
npx skills4sh remove agent-memory # uninstall agent-memory
npx skills4sh remove code-to-figma # uninstall code-to-figma
npx skills4sh remove discord-harvest # uninstall discord-harvest
npx skills4sh remove eleventy-nunjucks # uninstall eleventy-nunjucks
npx skills4sh remove figma-to-code # uninstall figma-to-code
npx skills4sh remove localhost-screenshots # uninstall localhost-screenshots
npx skills4sh remove skill-architect # uninstall skill-architectDefaults to ~/.claude/skills/. Override with --dest <dir> to target ~/.cursor/skills/, ~/.agents/skills/, or any path. Requires Node 22+. skills4sh remove --all --yes is intentionally not listed: it wipes every skill in <dest>, not just those installed by this package.
Skill structure
Each skill follows the Agent Skills specification:
skills/<skill-name>/
├── SKILL.md # Required: metadata + instructions
├── references/ # Supporting documentation
└── assets/ # Optional icons or static assetsSecurity
See SECURITY.md for the full compliance mapping, vulnerability disclosure process, and expected findings. Per-skill security manifests live in .security/.
Security scanning
Security scans are pinned to guardskills 1.2.1. agent-memory, discord-harvest, figma-to-code, and skill-architect scan without overrides. code-to-figma, localhost-screenshots, and eleventy-nunjucks have documented false-positive findings from instructional CI/env/secret/grep snippets; the CI matrix only accepts those known findings when they match SECURITY.md § Expected Security Findings.
npx [email protected] add t4sh/skills4sh --skill agent-memory --dry-run;
npx [email protected] add t4sh/skills4sh --skill code-to-figma --dry-run --force;
npx [email protected] add t4sh/skills4sh --skill discord-harvest --dry-run;
npx [email protected] add t4sh/skills4sh --skill eleventy-nunjucks --dry-run --force;
npx [email protected] add t4sh/skills4sh --skill figma-to-code --dry-run;
npx [email protected] add t4sh/skills4sh --skill localhost-screenshots --dry-run;
npx [email protected] add t4sh/skills4sh --skill skill-architect --dry-run;Skills do not auto-execute install or runtime code. Some skills may ship optional helper scripts under assets/; treat those as inert files unless a user or agent explicitly runs them.
License
MIT - see LICENSE for Details.
