The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the 'tai tool' bash command to list secret metadata and read decrypted secret values from the local environment. This is a legitimate use of command execution to interface with the vendor's secret storage system.
[INDIRECT_PROMPT_INJECTION]: The skill provides the agent with the ability to programmatically read sensitive credentials, creating an attack surface for manipulation via untrusted data. 1. Ingestion points: The agent may invoke the tool based on requests from processed data. 2. Boundary markers: Explicit instructions are included to never log or expose secrets in output. 3. Capability inventory: Uses shell execution of 'tai tool'. 4. Sanitization: The skill relies on agent instructions to maintain data confidentiality.

yao-secret