agent-pulse

Pass

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent to install the 'agentpulse-cli' package from PyPI using pip if the tool is not already available in the environment.- [COMMAND_EXECUTION]: The skill utilizes the 'agent-pulse' CLI and a custom Python script ('scripts/run_agent_pulse_snapshot.py') to execute various subcommands (e.g., status, doctor, models) via subprocess calls to gather system and session metrics.- [PROMPT_INJECTION]: The skill constitutes an indirect prompt injection surface as it processes and summarizes logs generated by external AI agents like Claude, Aider, and Cursor.\n
  • Ingestion points: Data is ingested through CLI output and aggregated JSON snapshots containing logs from multiple local agents.\n
  • Boundary markers: There are no explicit markers or instructions provided to the agent to treat the processed log content as untrusted or to ignore embedded instructions.\n
  • Capability inventory: The skill maintains capabilities to execute shell subcommands and configure environment variables for character encoding.\n
  • Sanitization: Although output is processed as JSON, the textual content from logs is not sanitized for potential natural language instruction overrides.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 7, 2026, 11:23 PM
Security Audit — agent-trust-hub — agent-pulse