The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the use of the --on-success CLI flag and build hooks (e.g., build:done), which allow the execution of arbitrary shell commands or JavaScript code after a build process completes.
Evidence in references/config.md: --on-success "echo Done!"
Evidence in references/features.md: tsdown --watch --on-success "node dist/index.mjs"
Evidence in references/advanced.md: hooks: { 'build:done': async (context) => { ... } }
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and execute packages using npx and to install various development dependencies from the npm registry. These resources include vendor-specific tools and well-known library plugins.
Evidence in SKILL.md: npm i -D tsdown typescript
Evidence in references/advanced.md: npx tsdown-migrate
Evidence in references/features.md: npx create-tsdown@latest, pnpm add -D unplugin-lightningcss, @rollup/plugin-babel, babel-plugin-react-compiler, etc.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process project configuration files (tsdown.config.ts, package.json) and project source code. These files are untrusted external inputs that could contain malicious instructions if the project being built is compromised.
Ingestion points: tsdown.config.ts, package.json, and source files specified in the entry field.
Boundary markers: None mentioned for the agent to distinguish between its instructions and the content of the config files.
Capability inventory: The skill can execute shell commands via on-success and hooks, perform file system writes for the build output, and run arbitrary JavaScript through the plugin system.
Sanitization: None mentioned for the interpolation of configuration values or file content.

tsdown