The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions for the manual installation of the 'ak' tool by cloning a repository and executing a shell script: git clone --branch v0.5.0 https://github.com/DailybotHQ/coding-agents-kit && ./coding-agents-kit/install.sh. Running unverified installation scripts from the internet is a significant security risk.
[EXTERNAL_DOWNLOADS]: The skill utilizes npx and git clone to fetch external resources from GitHub. Additionally, the ak install command is used to download and install third-party coding agent CLIs from vendor channels.
[CREDENTIALS_UNSAFE]: The skill manages a wide range of sensitive provider API keys (e.g., ZAI_CODING_API_KEY, XAI_API_KEY, AZURE_OPENAI_API_KEY, GEMINI_API_KEY). Although the instructions explicitly forbid the agent from printing or displaying these values, the tool provides the infrastructure to access them via environment variables and configuration files (~/.config/agentkit/env).
[INDIRECT_PROMPT_INJECTION]: The skill processes data returned from other agents via the ak run command.
Ingestion points: External data enters the agent context through the result_text and transcript fields in the JSON output of the ak run command.
Boundary markers: The skill includes a 'Trust boundary' section that explicitly instructs the agent to treat returned text as information to evaluate, not as instructions to execute.
Capability inventory: The skill has access to subprocess execution for managing profiles, installing software, and launching other agents.
Sanitization: The documentation advises the agent to verify the results of external agent runs with its own checks before relying on them.
[COMMAND_EXECUTION]: The skill relies on the execution of the ak CLI to manage accounts, configure environments, and launch interactive or headless agent sessions, which involves significant interaction with the user's host system.

agentkit