find-qualified-titles
Warn
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to download and install a package named 'deepline' from an external source.
- [REMOTE_CODE_EXECUTION]: The installation instructions utilize a custom, non-standard NPM registry ('https://code.deepline.com/api/v2/npm/') instead of the official public registry. This allows the execution of remote code from an unverifiable source that has not undergone standard public registry security scanning.
- [COMMAND_EXECUTION]: The skill includes shell commands for the agent to execute, including automated authentication workflows ('deepline auth register') and system configuration ('npm config set prefix').
- [PERSISTENCE]: The 'Quick Start' section includes commands that modify the system's 'PATH' environment variable ('export PATH="$HOME/.local/bin:$PATH"'), which alters command resolution for the current and potentially future sessions.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect instructions contained within external data.
- Ingestion points: User-supplied company domains and ICP (Ideal Customer Profile) filters processed by the 'find-qualified-titles' recipe.
- Boundary markers: No specific boundary delimiters or safety markers are used to separate untrusted search data from internal instructions.
- Capability inventory: The skill environment allows shell command execution, file system modification (mkdir), and network communication via the 'deepline' tool.
- Sanitization: There is no evidence of data sanitization or validation of the company-related data before it is used to drive discovery processes.
Audit Metadata