strategic-narrative-designer

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill manages untrusted external data sources such as scraped competitor pages and user-provided notes, which presents a surface for indirect prompt injection.\n
  • Ingestion points: External content enters the context via user pastes or research connectors like tavily.py and wayback.py (referenced in SKILL.md).\n
  • Boundary markers: The instructions explicitly direct the agent to treat this data as untrusted per SECURITY.md and to disregard any instructions embedded within the ingested content.\n
  • Capability inventory: The agent can read and write to the local memory/ directory and execute research scripts.\n
  • Sanitization: The skill relies on instructional guardrails to prevent the agent from executing commands hidden in external data.\n- [COMMAND_EXECUTION]: The skill utilizes local project scripts (e.g., registry-events.py) to handle state changes via a structured proposal mechanism. This pattern avoids direct manipulation of canonical files and routes all updates through a controlled execution flow.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 07:29 PM
Security Audit — agent-trust-hub — strategic-narrative-designer