writing-plans
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted project specifications to generate implementation plans that contain code and shell commands, creating a surface for indirect prompt injection.
- Ingestion points: The skill takes a "spec or requirements for a multi-step task" as input in
SKILL.md. - Boundary markers: There are no explicit delimiters defined to separate the untrusted specification text from the skill's instructions.
- Capability inventory: The resulting plans include shell commands (e.g.,
pytest,git) and implementation code that the agent is expected to execute in subsequent steps. - Sanitization: The skill employs a
plan-document-reviewersubagent to verify "Spec Alignment" and "scope creep," which helps mitigate the risk of malicious instructions being incorporated into the plan. - [COMMAND_EXECUTION]: The skill generates specific shell commands for inclusion in the implementation plan, such as
pytestfor testing andgitfor commits. While these are standard tools, the parameters and context of these commands are derived from the external, user-supplied specification.
Audit Metadata