The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes content from external websites using tools like WebFetch. This exposes the agent to potential instructions hidden within documentation pages that could override agent behavior.
Ingestion points: External URLs accessed via WebFetch and search results from WebSearch as described in SKILL.md.
Boundary markers: The instructions do not define clear delimiters or warnings to treat external content as untrusted data rather than instructions.
Capability inventory: The skill allows web browsing, search, and reading local file content within gems.
Sanitization: There is no explicit requirement for the agent to sanitize or filter out executable instructions from the documentation it summarizes.
[COMMAND_EXECUTION]: The skill involves running shell commands to facilitate source code exploration.
Evidence: The instruction in SKILL.md to use bundle show <gem_name> to locate installed Ruby gems on the local filesystem.
[EXTERNAL_DOWNLOADS]: The skill fetches documentation from external repositories and websites.
Evidence: The skill fetches configuration and guides from official documentation sites for Tailwind CSS, Ruby on Rails, and ReductoAI.

docs-researcher