The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from other local skills' SKILL.md files to identify documentation modules to fetch and save. This creates a potential vulnerability where data from an untrusted or compromised skill could influence the file operations of this maintenance tool.
Ingestion points: Reads SKILL.md files from subdirectories within ~/.claude/skills/ using the Read tool as described in SKILL.md.
Boundary markers: The instructions lack specific boundary markers or "ignore embedded instructions" warnings when handling the parsed reference lines.
Capability inventory: The skill uses Write, Edit, agent-browser, and WebFetch tools to modify the local filesystem and perform network operations.
Sanitization: The skill lacks explicit sanitization or validation of the module and crate_name variables extracted from external files before they are interpolated into filesystem paths for Write and Edit operations.
[EXTERNAL_DOWNLOADS]: Fetches documentation content from docs.rs. This is a well-known service for hosting Rust documentation, and the download activity is consistent with the skill's primary purpose.
[COMMAND_EXECUTION]: Utilizes shell commands including ls and bash loops to scan directories and verify the existence of documentation files within the agent's environment.

core-fix-skill-docs