asu-recap
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of AI programming conversations and project delivery records (SKILL.md). This creates a surface for indirect prompt injection where instructions hidden in the processed data could attempt to influence the agent's behavior. However, the skill does not possess dangerous capabilities like command execution or network access, and it includes explicit instructions for data sanitization and privacy masking.
- Ingestion points: Data enters the context via user-provided conversation logs and project materials (SKILL.md).
- Boundary markers: The instructions do not define specific delimiters for the input data, though they provide a structured workflow for handling the content.
- Capability inventory: The skill's actions are limited to generating structured text output; no subprocess calls, file writes, or network operations are present.
- Sanitization: Step 2 of the workflow and the '表达边界' section explicitly mandate the generalization or replacement of sensitive data including tokens, keys, emails, and internal paths.
- [SAFE]: The skill includes robust privacy protection instructions, requiring the agent to perform privacy checks and replace sensitive identifiers (tokens, keys, client IDs) with generalized placeholders (SKILL.md).
- [SAFE]: No external dependencies, remote code downloads, or unauthorized command execution patterns were detected in the provided files.
Audit Metadata