The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions in SKILL.md and related documentation focus on implementation details for webhook handlers and cryptographic verification. No instructions attempting to override agent behavior, bypass safety filters, or extract system prompts were found.
[CREDENTIALS_UNSAFE]: The skill uses environment variables (TOKEN_WEBHOOK_PUBLIC_KEY) for configuration. The documentation correctly identifies this as a public key and not a shared secret. Example .env files use safe placeholders like your_member_ed25519_public_key_base64url.
[EXTERNAL_DOWNLOADS]: The skill references npx hookdeck-cli for local development. This is a well-known tool provided by the skill's author (hookdeck) and is hosted on the standard npm registry. Dependencies listed in package.json and requirements.txt (such as express, fastapi, and cryptography) are standard, well-known libraries for the intended functionality.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting external data via HTTP POST requests in the provided examples (examples/express/src/index.js, examples/fastapi/main.py, examples/nextjs/app/webhooks/tokenio/route.ts). However, it implements strong mitigation by mandating Ed25519 signature verification on the raw body before JSON parsing or logic execution, effectively preventing the processing of unauthorized or tampered data.
[COMMAND_EXECUTION]: All scripts included in the examples are standard web server implementations. They do not contain any eval(), exec(), or subprocess calls that use external data or perform dangerous operations.

tokenio-webhooks