arcgis-exb-widget-dev
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection because it processes untrusted user requirements to perform file system modifications and command execution.
- Ingestion points: User-provided Requirement Payloads described in README.md and the Requirement Clarification section of SKILL.md.
- Boundary markers: The instructions in SKILL.md and related references do not specify the use of delimiters or ignore embedded instructions warnings when the agent processes external requirements.
- Capability inventory: The skill grants the agent the ability to write multiple files, including source code, manifests, and configs as defined in references/manifest-and-files.md, and execute shell commands such as npm run tscheck mentioned in SKILL.md.
- Sanitization: No sanitization, escaping, or filtering of the requirement payload content is specified in SKILL.md or the development workflow.
Audit Metadata