research
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from project files such as
wayfinder.md,decision-log.md, anddomain-model.mdto formulate research hypotheses and code spikes. There are no mentions of boundary markers or sanitization techniques to prevent malicious instructions embedded in these files from influencing the agent's behavior. - Ingestion points:
wayfinder.md,decision-log.md,domain-model.md,package.json, and other source code files. - Boundary markers: Absent.
- Capability inventory: Writing and executing scripts in the
scratch/directory, capturing logs, and making implementation recommendations. - Sanitization: Absent.
- [DYNAMIC_EXECUTION]: The skill is designed to create and run temporary research scripts (e.g.,
spike_api_test.tsor.py) in a dedicated workspace directory. This involves the generation and execution of code at runtime based on the technical unknowns it identifies. - [COMMAND_EXECUTION]: The procedure involves running benchmark tests and API integration probes, which likely involves shell command execution to perform performance measurements (latency, throughput, memory footprint).
Audit Metadata