The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell commands to manage the onboarding process. This includes using npx for updates, uv to run a validation script, and the calibrate CLI for agent management, configuration, and authentication.
[EXTERNAL_DOWNLOADS]: The skill fetches updates for its own instructions and scripts using the npx skills update command. It also provides instructions for the user to install the vendor's CLI tool via Homebrew (brew install dalmia/tap/calibrate).
[DYNAMIC_EXECUTION]: Runtime execution of code occurs through the self-update mechanism and the execution of a bundled Python script (scripts/validate.py). The validation script is limited to the Python standard library and performs static checks on a local state file to ensure the onboarding process is complete.
[INDIRECT_PROMPT_INJECTION]: The skill maintains project state in .calibrate/onboard.md, which it reads to resume onboarding sessions. This creates a surface where external data (the state file) influences the agent's instructions.
Ingestion points: Reads state from .calibrate/onboard.md in SKILL.md and scripts/validate.py.
Boundary markers: None explicitly defined for the state file content, though the agent is instructed to follow the structure of the markdown file.
Capability inventory: Executes shell commands (npx, uv, calibrate) and performs file system operations (mkdir, cp).
Sanitization: The validate.py script uses regular expressions to filter out markers and italicized text before processing, acting as a basic validator for the ingested state.

onboard