skill-discovery
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill retrieves the README.md from the L-yifan/skills repository on GitHub to build its discovery list.
- [COMMAND_EXECUTION]: The skill identifies and extracts installation command strings (e.g., npx commands) from remote documentation.
- [DATA_EXFILTRATION]: The skill reads local project files like AGENTS.md and CLAUDE.md to derive keywords for remote searching; access is limited to a maximum of three files and 12 KB per file.
- [PROMPT_INJECTION]: The skill ingests untrusted data from an external repository which is a potential surface for indirect prompt injection. Ingestion points: The remote README.md from the L-yifan/skills repository. Boundary markers (absent): The skill instructions do not include specific delimiters or warnings for the agent to ignore content in the README. Capability inventory: Metadata file reading and outputting actionable commands for user execution. Sanitization (absent): The skill extracts and displays strings directly from the source repository.
Audit Metadata