The Agent Skills Directory

[DATA_EXFILTRATION]: The skill facilitates network communication with iGrant.io's official API endpoints (demo-api.igrant.io and staging-api.igrant.io) for proxying verifiable credential operations. These operations are restricted to specific allow-listed paths (v2/config/digital-wallet/openid/sdjwt/credential/*) to maintain a least-privilege security posture and prevent unauthorized data transmission to external domains.
[COMMAND_EXECUTION]: The package includes standard scripts for development and webhook registration using tsx to execute TypeScript files. These scripts are intended for local administrative use during setup and do not accept unvalidated remote input for shell execution.
[PROMPT_INJECTION]: The skill implementation includes an indirect prompt injection surface by processing external webhook payloads and proxying browser requests. It mitigates this risk by requiring and validating X-iGrant-Signature HMAC headers using a shared secret and restricting proxy access to specific OWS path prefixes, ensuring only authenticated and authorized data flows through the system.
[CREDENTIALS_UNSAFE]: The implementation follows security best practices for secret management by resolving organization-specific API keys from environment variables (OWS_TENANT_<SLUG>_API_KEY) instead of hardcoding sensitive strings. It ensures these credentials are only used for server-to-server communication and are never exposed to the client-side.

igrantio-issuer-backend