dev-swarm-stage-tech-research
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is instructed to install dependencies and execute automated tests as part of technical research projects (Step 4.1). This includes running test frameworks like pytest, jest, and playwright on newly generated code.- [EXTERNAL_DOWNLOADS]: The skill fetches documentation from the internet (Step 2.1) and suggests installing third-party packages and tools such as pnpm, uv, and cmake to validate technical assumptions.- [PROMPT_INJECTION]: The skill has a surface for Indirect Prompt Injection because it ingests untrusted data from external internet documentation and previous project stages (00-init-ideas through 03-mvp) which is then used to inform code generation and execution.
- Ingestion points: Markdown files in folders 00 through 03 and external documentation fetched from the internet.
- Boundary markers: None specified in the instructions to separate untrusted data from system instructions.
- Capability inventory: File system writing (Step 3.1), dependency installation, and command execution (Step 4.1).
- Sanitization: No explicit sanitization or validation of the fetched documentation is required before it influences code implementation.
Audit Metadata