The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing various system commands and local shell scripts via the deno task command, including scripts/check-local-dev.sh and scripts/restart-local-dev.sh.
[REMOTE_CODE_EXECUTION]: The ct check and ct piece new commands compile and evaluate local TypeScript and JSX code (pattern.tsx), performing dynamic code execution at runtime.
[CREDENTIALS_UNSAFE]: The skill workflow involves creating and managing an identity key (claude.key) stored in the local directory and utilizing environment variables (CT_IDENTITY) for authentication.
[DATA_EXFILTRATION]: The CLI is configured to communicate with external API endpoints, such as https://toolshed.saga-castor.ts.net/, establishing a network channel for potentially sensitive data transfer.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates reading and inspecting state from "pieces" and remote patterns.
Ingestion points: Data is ingested via piece get, piece inspect, and the evaluation of pattern.tsx files.
Boundary markers: None identified in the prompt templates.
Capability inventory: The skill has the capability to execute subprocesses and compile code via deno task.
Sanitization: There is no evidence of sanitization or validation of the data retrieved from external pieces before it is processed or presented to the agent.