barcode-capture-capacitor
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and modify untrusted data from the user's project files, such as
package.jsonand application source code. This represents a standard attack surface for coding-assistant skills where content in project files could influence agent behavior. - Ingestion points:
package.json,package-lock.json, and application source files (e.g.,app.js) as described inreferences/migration.mdandreferences/integration.md. - Boundary markers: Not present.
- Capability inventory: File read/write, network fetching of documentation.
- Sanitization: Not present.
- [PROMPT_INJECTION]: The skill contains instructions such as "Do not tell the user to check the docs themselves" and "Do not just show the code in chat; apply it to the file." These are evaluated as benign instructions aimed at ensuring agent helpfulness and directing the use of file-editing tools, rather than attempts to conceal malicious activity.
Audit Metadata