The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through the ingestion and interpolation of untrusted exchange identifiers into API request paths.\n
Ingestion points: Identifier strings such as exchangeId and credentialExchangeId are accepted by methods in owsClient.ts and sseClient.ts.\n
Boundary markers: The code lacks sanitization or explicit delimiters to prevent path-traversal sequences (e.g., ../) in these identifiers.\n
Capability inventory: The library utilizes fetch (including DELETE operations) and the EventSource API across owsClient.ts and sseClient.ts.\n
Sanitization: Identifiers are directly interpolated into template literals (e.g., ${base}/${id} for deletion in sseClient.ts), potentially allowing access to unintended API paths if the identifier is maliciously crafted.\n- [COMMAND_EXECUTION]: The openInWallet helper in QrCode.tsx assigns a provided URI directly to window.location.href. The absence of URI scheme validation could allow for the execution of javascript: URIs if the URI is sourced from untrusted data.\n- [EXTERNAL_DOWNLOADS]: The skill specifies qrcode and react as dependencies in package.json, which are standard and necessary for its functionality as a React-based frontend client.

igrantio-frontend-client