orient

Warn

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses the uvx utility to download the third-party showboat package from the official Python Package Index (PyPI).
  • [REMOTE_CODE_EXECUTION]: The showboat package is executed at runtime to manage the orientation document lifecycle, including initialization and verification.
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute various system commands such as find, git, and uvx, and employs the Write tool to create local directories and documentation files.
  • [PROMPT_INJECTION]: The skill reads and processes untrusted repository content, presenting an indirect prompt injection surface. Ingestion points: Repository manifest files, READMEs, and source code modules read during exploration. Boundary markers: No specific boundary markers or instructions to ignore embedded commands are present. Capability inventory: Access to Bash, Write, and uvx allows the agent to execute commands and modify the local filesystem based on processed content. Sanitization: No evidence of sanitization or validation of the ingested repository content is provided before it is used for synthesis or command planning.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 8, 2026, 03:37 AM