The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The documentation includes illustrative example credentials using common API key prefixes. Evidence: Example OpenAI-style keys (sk-123, sk-abc123) and a Telegram-style token (123456:ABC) are featured in the CLI command examples in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes agent information which is externally defined and could contain instructions targeting the LLM. 1. Ingestion points: Agent names, machine IDs, and status metadata retrieved via blink_agent_list and blink_agent_status. 2. Boundary markers: No explicit delimiters or instructions to ignore embedded content within agent metadata are present. 3. Capability inventory: The skill has the capability to write and delete secrets using blink_agent_secrets_set and blink_agent_secrets_delete. 4. Sanitization: No evidence of validation or sanitization of agent-defined metadata was found.
[COMMAND_EXECUTION]: The skill documents and utilizes the blink CLI tool to perform agent management and secret configuration. Evidence: Use of blink agent and blink secrets commands to interact with the Blink Claw hosting platform.

blink-agents