deco-full-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches a validation script from https://deco.cx/validate. This is the official validation tool for the Deco e-commerce platform that the skill is designed to analyze.
- [REMOTE_CODE_EXECUTION]: Executes the remote validation script using 'deno run -A'. The '-A' flag grants full permissions (network, file, env access) to the script, which is necessary for it to validate local TypeScript Props against JSON block configurations. This represents an intended use of official diagnostic tooling.
- [COMMAND_EXECUTION]: Uses system commands such as 'git log' to identify code owners and top committers, and 'grep' to search for configuration patterns in project files. These commands are executed locally for data gathering during site onboarding.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code and configuration files to generate documentation. * Ingestion points: Reads project files including 'deno.json', '.deco/blocks/*.json', and files in the 'sections/', 'loaders/', and 'actions/' directories. * Boundary markers: No specific boundary markers or instructions to ignore embedded prompts are defined for the ingested data. * Capability inventory: Executes 'git log' and 'deno run -A', and performs file system operations to generate 'AGENTS.md' and other reports. * Sanitization: No explicit sanitization or validation of the processed code content is mentioned before it is included in documentation. The risk is minimized as the skill is focused on static analysis and generating reports rather than making automated decisions based on the content.
Audit Metadata