core-fix-skill-docs
Pass
Audited by Gen Agent Trust Hub on Jun 25, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it parses SKILL.md files from the ~/.claude/skills/ directory to extract metadata. This data is then used to construct shell commands and sub-agent prompts. A malicious skill file could inject crafted strings to manipulate the agent's behavior. 1. Ingestion points: Reads SKILL.md files located in the local skills directory. 2. Boundary markers: No explicit delimiters or instructions are used to separate the external data from the skill's logic. 3. Capability inventory: Uses agent-browser CLI, WebFetch, file system writes, and sub-agent task generation. 4. Sanitization: No validation or escaping is applied to the crate_name or module variables before they are interpolated into commands.
- [COMMAND_EXECUTION]: The skill executes shell commands and uses the agent-browser CLI. Because parameters like crate_name are derived from external files, there is a risk of command injection if those files contain malicious metadata.
- [EXTERNAL_DOWNLOADS]: Fetches documentation from docs.rs, which is a well-known and official service for the Rust programming language community.
Audit Metadata