widget-migrate

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill employs Bash scripts and command-line utilities such as awk, git, and glob to automate the detection of widget types and the resolution of source code paths.
  • Evidence: Scripts in operations/detect-widget-type.md and operations/resolve-widget-source.md use shell commands to inspect the filesystem and verify repository availability.
  • [EXTERNAL_DOWNLOADS]: The skill performs repository cloning from an internal GitHub Enterprise instance to ensure that the required legacy source code is available for analysis.
  • Evidence: operations/resolve-widget-source.md includes logic to execute 'git clone' from internal domains like ghe.coxautoinc.com.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests and processes legacy source code to generate high-level technical specifications and implementation stories.
  • Ingestion points: The skill reads Velocity templates (.vm), Groovy scripts, and XML/JSON configuration files (e.g., portlet-portlets.xml, prefs.json) from the local filesystem.
  • Boundary markers: The pipeline does not specify the use of clear delimiters or instruction-ignore warnings when processing the content of these files for output generation.
  • Capability inventory: The skill has the capability to write and edit files (allowed-tools: Write, Edit), which are used to produce implementation epics and specs based on the analyzed data.
  • Sanitization: No explicit sanitization or filtering mechanisms are described to prevent instructions embedded within legacy code from influencing the agent's migration output.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 05:28 PM
Security Audit — agent-trust-hub — widget-migrate