The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local shell scripts included in the package, specifically scripts/new-canvas and scripts/tq, to manage the tldraw application lifecycle and perform API requests.
[DATA_EXFILTRATION]: The skill accesses sensitive local configuration data to retrieve security credentials.
Evidence: The scripts/tq helper and references/shared/bridge.md documentation describe reading server.json from sensitive paths (e.g., ~/Library/Application Support/tldraw/server.json) to extract an authorization token and port. While intended for local communication with localhost, this pattern involves handling active authentication tokens.
[REMOTE_CODE_EXECUTION]: The skill facilitates the execution of dynamically generated code within an external application environment.
Evidence: The skill provides instructions for sending JavaScript payloads to the /api/doc/:id/exec and /api/doc/:id/script-workspace endpoints to manipulate the tldraw canvas and register durable document scripts.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its data processing workflow.
Ingestion points: The agent is instructed to read repository source code and user-provided materials to generate factual whitepaper content (SKILL.md).
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when processing this external content.
Capability inventory: The agent has access to shell script execution and a local API capable of running JavaScript within the tldraw application.
Sanitization: There is no evidence of sanitization or filtering of the ingested repository data before it enters the agent's context.

whitepaper