make-the-docs-trustworthy
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted documentation content and is explicitly instructed to execute commands or follow paths found within those documents.
- Ingestion points: The agent reads and analyzes documentation files (e.g., README.md, docs/*.md) from the repository to verify their accuracy.
- Boundary markers: There are no specific instructions or delimiters provided to isolate the documentation content or to warn the agent about potentially malicious instructions embedded within the text.
- Capability inventory: The
SKILL.mdfile commands the agent to "Run the command the page describes", "Open the path it names", and "Follow the link and see where it lands", providing file system, command execution, and network access capabilities. - Sanitization: The skill lacks requirements for sanitizing or validating the commands, paths, or URLs extracted from the documentation before they are acted upon.
- [COMMAND_EXECUTION]: The skill's primary verification mechanism involves the agent executing arbitrary shell commands found within documentation files to confirm their validity. This design, while functional for the skill's purpose, allows for the execution of potentially dangerous commands if the documentation is compromised.
Audit Metadata