skills/howells/arc/browse/Gen Agent Trust Hub

browse

Warn

Audited by Gen Agent Trust Hub on May 10, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill is configured to locate and read sensitive environment files such as .env, .env.local, and .env.development to determine the application port during its context scan phase. Accessing these files represents a potential exposure risk for any credentials or secrets stored within the environment configuration.
  • [PROMPT_INJECTION]: The skill operates on untrusted data retrieved from a web application, creating an attack surface for indirect prompt injection where malicious content in the app could influence the agent's behavior.
  • Ingestion points: Data is ingested from the web application via browser snapshot, page reading, and console log tools as described in the browser setup and session phases.
  • Boundary markers: The instructions lack clear delimiters or specific guidance to the agent to treat the retrieved content as untrusted or to disregard potential instructions embedded within the application interface.
  • Capability inventory: The agent has permissions to read local files, execute git commands, and perform browser automation with interaction capabilities.
  • Sanitization: No sanitization or validation of external web content is specified before the agent processes and evaluates it.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 10, 2026, 04:39 PM
Security Audit — agent-trust-hub — browse