The Agent Skills Directory

[SAFE]: The skill's primary content consists of technical documentation and configuration templates. No malicious instructions, obfuscation, or persistence mechanisms were detected. Behavior aligns with the stated purpose of a developer guide.
[EXTERNAL_DOWNLOADS]: The skill includes instructions to fetch and install various packages and CLI tools from official registries and well-known technology domains.
Evidence:
Commands to add dependencies like @vitejs/plugin-react, hono, vitest, and biome via pnpm or npm.
Usage of pnpm dlx shadcn@latest init to initialize UI components.
[COMMAND_EXECUTION]: Provides numerous commands for the agent to initialize projects, run linting/formatting, and execute tests.
Evidence:
biome check --write . for code maintenance.
vitest run for executing the test suite.
shadcn add for component management.
[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection by suggesting the agent fetch remote documentation for extended context.
Ingestion points: references/hono.md suggests fetching context from https://hono.dev/llms-full.txt.
Boundary markers: None specified for the remote text file.
Capability inventory: The skill encourages execution of shell commands (pnpm, npm) and file system operations (writing source code and configuration files).
Sanitization: No sanitization or validation of the remote documentation is specified.
Contextual Note: This is a standard practice for providing LLMs with up-to-date documentation from official sources.

typescript-dev