pulumi-modify-existing
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes existing infrastructure code (e.g.,
__main__.py) which represents an untrusted data ingestion surface. Malicious instructions embedded in these files (via comments or string literals) could attempt to influence the agent's behavior during the editing process. - Ingestion points: Existing codebase files like
__main__.pymentioned in SKILL.md. - Boundary markers: Absent; there are no specific instructions to delimit data from instructions or ignore embedded content.
- Capability inventory: Shell command execution via
pulumi preview. - Sanitization: No explicit sanitization or filtering of external code content is performed.
- [COMMAND_EXECUTION]: The skill directs the agent to execute
pulumi preview --stack <stack-name>to validate changes. While this is the intended primary purpose for verifying infrastructure plans, it involves direct shell interaction and command execution. - [DYNAMIC_EXECUTION]: Pulumi operations involve executing the underlying infrastructure program (written in languages like Python or TypeScript). The agent executes code that it has potentially modified or code that exists within the stack's directory to generate the preview.
Audit Metadata