The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (scripts/tq and scripts/new-canvas) to interact with the tldraw application and manage filesystem operations. These scripts use system tools like curl and python3 to communicate with the application's local API.
[DATA_EXFILTRATION]: The scripts/tq script reads the application's local configuration file (server.json) to retrieve the port and authorization token required for API access. This data access is restricted to the local machine for legitimate inter-process communication with the tldraw app.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes user-provided prose or transition lists in SKILL.md. Evidence chain:
Ingestion points: User-provided prose or transition lists (Step 1 of Workflow).
Boundary markers: Absent; the instructions do not specify delimiters for the input data.
Capability inventory: Execution of shell scripts and local API calls via scripts/tq.
Sanitization: Absent; while the skill performs normalization, it lacks explicit logic to filter or escape embedded instructions.
[SAFE]: The binary file assets/blank.tldraw is a standard ZIP-compressed archive used as a template by the tldraw application. Analysis of its contents (including db.sqlite and metadata.json) confirms it is a legitimate asset for the skill's stated purpose.

state-diagram