ae-project-semantic
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from 'project asset packages' to generate business recommendations, creating a potential surface for indirect prompt injection.
- Ingestion points: SKILL.md (Step 6) identifies several ingestion points including
manifest.json,catalog/*.jsonl,indexes/*.jsonl, and asset details. - Boundary markers: The instructions do not define explicit delimiters or warnings to the agent to ignore or isolate natural language instructions that might be embedded within the asset metadata, display names, or descriptions.
- Capability inventory: The agent has the ability to execute
ae-clicommands that modify the state of the project, includingcandidate submit,candidate enable, andproject-semantic delete. - Sanitization: No sanitization or validation of the natural language content within the ingested files is mentioned before the agent performs 'business topic discovery' and 'evidence interpretation'.
Audit Metadata