bfs-lobster-agent-creator

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the date command to generate timestamps for file versioning and grep to ensure all template placeholders have been properly substituted. These are standard development tasks performed in the local environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as a text processor that interpolates user-supplied role mandates and behavioral guidelines into agent persona templates.
  • Ingestion points: User-provided inputs such as role_and_mandate, name, and tone_overrides provided via the input schema in SKILL.md.
  • Boundary markers: Uses Markdown headers and placeholder tags ({{...}}) for structure, though it does not provide explicit defensive filtering for the user's specific text inputs.
  • Capability inventory: Limited to file read/write operations and local execution of non-destructive shell utilities (date, grep).
  • Sanitization: Literal substitution is performed; no escaping of user input is implemented, which is typical for a persona authoring tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 03:45 PM
Security Audit — agent-trust-hub — bfs-lobster-agent-creator