The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the keyvault and keyvault-ai packages directly from an unverified GitHub repository (github.com/ylz201/keyvault.git). as this repository belongs to an unverified author, its source and contents are unverifiable.
[COMMAND_EXECUTION]: The skill utilizes pip install with remote Git URLs for installation. It also includes the keyvault inject command for executing local scripts with environment variable injection and provides an MCP server setup via python -m keyvault.mcp_server.
[CREDENTIALS_UNSAFE]: The tool manages and accesses sensitive files at ~/.keyvault/master.key and ~/.keyvault/vault.db to store encryption keys and secrets. While associated with the skill's purpose, these represent access to sensitive local security infrastructure.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by processing external secret data that is later injected into process environments.
Ingestion points: Secret values and keys imported from .env files or CLI arguments.
Boundary markers: No delimiters or instructions to ignore embedded commands are specified for the processed data.
Capability inventory: File system access to the user's home directory and subprocess spawning via keyvault inject.
Sanitization: No input validation or sanitization of stored secret values is described.

secrets-manager