The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes local CLI tools such as git, gh (GitHub CLI), and node to manage repository state and fetch metadata. These operations are restricted to the local environment and well-known services.
[CREDENTIALS_UNSAFE]: The skill manages Linear API keys through linear-graphql.mjs. It follows security best practices by using the macOS Keychain (security utility) for encrypted storage rather than saving secrets in plain text. It also uses stty to hide user input during setup.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface when reading untrusted data from issue descriptions and pull request bodies. It implements a mitigation in workflow-contract.mjs (classifyInstructionSource) using regex patterns to detect and ignore attempts to override system policies or bypass security checks.
[DYNAMIC_EXECUTION]: The skill dynamically generates shell commands for the agent to execute. However, it employs a robust quote function in action-briefs.mjs to escape arguments, which effectively mitigates command injection risks from malicious ticket data.

ziw-orchestrate