install-anti-sycophancy

Pass

Audited by Gen Agent Trust Hub on Jun 28, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill configures a hook to inject behavioral instructions into the agent's context for every message, attempting to override default conversational styles.
  • [COMMAND_EXECUTION]: The skill generates and installs shell scripts and modifies the agent harness configuration to execute them automatically during active sessions.
  • [PROMPT_INJECTION]: The skill processes conversation transcripts through a monitoring hook, creating an indirect injection surface.
  • Ingestion points: Transcript file path provided by the harness.
  • Boundary markers: Absent.
  • Capability inventory: Uses shell utilities (tail, grep, jq) and local file writing.
  • Sanitization: Log output is structured via jq.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 28, 2026, 01:39 AM
Security Audit — agent-trust-hub — install-anti-sycophancy