The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill decodes QR codes from the device camera, which creates a surface for indirect prompt injection if the agent subsequently processes the decoded text as instructions.
Ingestion points: Untrusted data enters the agent context via the qrResults array returned by the useQRScanner hook defined in SKILL.md.
Boundary markers: The implementation lacks explicit delimiters or instructions to the agent to ignore or isolate commands that might be embedded within the scanned QR data.
Capability inventory: The decoded data is stored in memory and displayed in the frontend, making it accessible to the agent for further reasoning or task execution.
Sanitization: There is no evidence of validation, sanitization, or filtering of the decoded QR content before it is added to the results history.
[EXTERNAL_DOWNLOADS]: The skill configuration allows for loading the jsQR decoding library from an external URL, defaulting to the jsdelivr CDN.
Evidence: The QRScannerConfig interface in SKILL.md includes a jsQRUrl property which defaults to a public CDN for loading the decoding logic.

extension-qr-code