writing-plans
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external, untrusted specifications to generate structured implementation plans without safety constraints.
- Ingestion points: User-provided specifications or requirements provided to the agent at runtime, as referenced in the skill's description.
- Boundary markers: Absent. The instructions do not define delimiters or directives to isolate or ignore instructions embedded within the source requirements.
- Capability inventory: The skill facilitates the generation of files and shell commands (e.g.,
git,pytest) intended for execution by subagents or parallel sessions. - Sanitization: Absent. There is no logic provided to escape or validate content before it is interpolated into implementation plans.
Audit Metadata