deco-full-analysis

Pass

Audited by Gen Agent Trust Hub on Mar 11, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill executes a remote validation script from the official Deco platform (https://deco.cx/validate) using 'deno run -A' to verify site configurations. \n- [COMMAND_EXECUTION]: The skill uses 'git log' to extract contributor information for generating code ownership reports. It also provides shell commands for file system auditing and instructions for removing unused platform-specific files via 'rm'. \n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted project files (JSON and TSX) to generate its reports. \n
  • Ingestion points: reads project metadata (deno.json), page configurations (.deco/blocks/*.json), and custom component source code (sections/, loaders/, actions/). \n
  • Boundary markers: Absent; the skill does not use specific delimiters to isolate site content from its analytical instructions. \n
  • Capability inventory: File system access (read, delete), network operations (curl), and local command execution (git, deno). \n
  • Sanitization: Project files are analyzed directly without sanitization or escaping of their content.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 11, 2026, 11:22 PM