skill-from-book
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to use shell commands like
wc -landgrepto analyze the structure of the input book file. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external markdown files, creating a surface for indirect prompt injection attacks.
- Ingestion points: The content of the user-provided
book.mdis read by both the main agent (for TOC extraction) and subagents (for knowledge extraction). - Boundary markers: The skill lacks explicit delimiters or instructions to treat the book content as data only, potentially allowing embedded instructions in the book to influence the agent's behavior.
- Capability inventory: The skill environment allows for shell command execution, creation of subagents, and extensive file system write operations to generate the new skill structure.
- Sanitization: No sanitization, escaping, or validation logic is applied to the text extracted from the book before it is processed or written to new files.
Audit Metadata