brand-storytelling
Pass
Audited by Gen Agent Trust Hub on Feb 19, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE] (SAFE): The skill consists entirely of Markdown files (
SKILL.md,README.md, and several files inreferences/). There are no Python, JavaScript, or Shell scripts included, and no external dependencies are declared. - [DATA_EXFILTRATION] (SAFE): No network-capable commands (e.g.,
curl,wget) or patterns for accessing sensitive local configuration files (e.g.,~/.ssh,.env) were detected. - [REMOTE_CODE_EXECUTION] (SAFE): The skill does not perform any remote downloads or dynamic execution of external content.
- [PROMPT_INJECTION] (LOW): The skill is designed to process user-provided story material. While this constitutes an attack surface for indirect prompt injection, the skill includes explicit safety instructions to "refuse fabrication" and verify claims against provided proof. Furthermore, the skill has no executable capabilities that could be exploited via injection.
- Ingestion points: User input fields for "Company", "Audience", and "Raw story material" defined in
SKILL.md. - Boundary markers: Absent; the skill relies on the model's inherent ability to distinguish instructions from data.
- Capability inventory: None; the skill generates text only and does not contain code, subprocess calls, or file-system writing logic.
- Sanitization: None; the skill relies on the agent's base safety layers.
Audit Metadata