The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow for handling external data in the form of cryptographically signed verifiable credentials.
Ingestion points: The skill parses credential.credentialJson returned by the Android Credential Manager API as seen in SKILL.md and references/android/identity/digital-credentials/credential-verifier.md.
Boundary markers: The skill uses the OpenID4VP and SD-JWT protocols which have defined internal structures. Instructions warn developers to perform full validation on the server and use client-side parsing only for UI updates.
Capability inventory: The skill provides code to request and process identity attributes. It does not include direct shell execution or dynamic evaluation of this data within the agent's context, but it prepares the application environment to handle external tokens.
Sanitization: Documentation explicitly mandates server-side cryptographic validation, signature checking using official Google JWKs, and nonce validation to prevent replay attacks.
[EXTERNAL_DOWNLOADS]: The skill directs users to download dependencies from official repositories (e.g., androidx.credentials:credentials:1.7.0-alpha03). It also references standard specifications and documentation from well-known organizations such as W3C, IETF, and the OpenID Foundation. These are standard and safe development practices.
[COMMAND_EXECUTION]: The skill mentions standard development commands including android skills add verified-email for the Android CLI and keytool for identifying app signing certificate fingerprints. These are legitimate tools for the intended use case.

verified-email