architecture-aware-init
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a workflow where web search results are used to influence architectural decisions via an automated script.
- Ingestion points: In modules/research-flow.md, the results from WebSearch are synthesized and stored in research.json.
- Boundary markers: There are no explicit instructions or delimiters mentioned to prevent the agent from following malicious instructions hidden in the search results.
- Capability inventory: The skill possesses file system write capabilities (directory scaffolding in modules/scaffold-generation.md) and command execution capabilities (Python scripts in modules/script-integration.md).
- Sanitization: No sanitization or filtering logic is described for the data passed from the web search to the research.json file used by the scoring script.
- [COMMAND_EXECUTION]: The skill frequently invokes local Python scripts to automate architecture research and project scaffolding.
- Evidence: uv run python plugins/attune/scripts/attune_arch_init.py and uv run python plugins/attune/scripts/architecture_researcher.py are executed in modules/research-flow.md and modules/script-integration.md respectively.
- These scripts reside in the plugins/attune/ directory, which corresponds to the vendor's own tooling namespace.
Audit Metadata