implement
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies and processes local specification artifacts which represent a potential attack surface for indirect instructions.
- Ingestion points: The skill reads and depends on files such as
manifest.md,plan/README.md,requirements.md, andsolution.md. - Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore embedded commands within these artifacts.
- Capability inventory: The skill dispatches work to sub-skills (
implement-direct,implement-incremental,implement-factory) which likely possess capabilities for file writing and code execution. - Sanitization: No sanitization or validation of the content within the specification artifacts is specified before they are handed off to the execution sub-skills.
Audit Metadata