deco-cms-layout-caching
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill involves the agent reading and processing CMS block configuration files from the local filesystem (e.g.,
.deco/blocks/*.json). - Ingestion points: The
isRawSectionLayoutlogic processes thedecofileDatastructure, which contains configuration objects derived from local JSON files. - Boundary markers: No explicit delimiters or boundary markers are defined to isolate processed CMS data from agent instructions.
- Capability inventory: The skill implements CMS resolution and loader execution mechanisms. It also suggests the use of diagnostic shell commands for verifying block metadata.
- Sanitization: The implementation uses a depth-limited recursive search (5 levels) to prevent infinite loops during block resolution, though specific data sanitization for block properties is not detailed.
- [COMMAND_EXECUTION]: The documentation includes a diagnostic command example using
catandpython3 -cto parse JSON metadata. This is a standard troubleshooting utility provided for developer guidance and does not involve remote code execution or suspicious triggers.
Audit Metadata