stress-test
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to use the Bash tool to execute Proof of Concept (POC) code and perform directory cleanup using
rm -rf .poc-stress-test/in Phase 6. - [DYNAMIC_EXECUTION]: In Phase 5, the agent is instructed to write and execute small representative tests (POCs) based on the assumptions identified in the technical plan. This involves generating and running scripts at runtime.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external technical plans which may come from untrusted sources, creating a risk that malicious instructions within the plan could influence the agent's behavior.
- Ingestion points: Technical plans are read from the conversation history during Phase 1 ("Extract & Decompose").
- Boundary markers: The instructions do not specify the use of delimiters or specific warnings to ignore instructions embedded within the technical plan.
- Capability inventory: The skill has access to powerful tools including
Bash,Write,Edit,WebFetch, andTask(for spawning sub-agents). - Sanitization: No explicit sanitization or validation of the input plan is performed before it is used to define the logic for POC execution.
Audit Metadata