ae-project-semantic
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project asset packages containing data that is not directly controlled by the skill, such as dashboard notes and report descriptions, creating a surface for indirect instructions.\n
- Ingestion points: The skill reads multiple metadata and data files, including catalog files (.jsonl), indexes (.jsonl), and detailed asset definitions in details/normalized/** and details/raw/** from exported packages.\n
- Boundary markers: While the skill instructions provide guidance on evidence interpretation, they lack explicit boundary markers or directives for the agent to ignore potentially malicious instructions embedded in asset notes or memos.\n
- Capability inventory: The skill possesses the capability to execute complex ae-cli commands for exporting, validating, and submitting project semantics, which could be influenced by injected content.\n
- Sanitization: There is no evidence of specific sanitization or filtering logic applied to the textual evidence sourced from assets before it is incorporated into the generated semantic candidates.
Audit Metadata