The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates on documents and data sources that could potentially contain hidden instructions designed to influence the agent's behavior during processing or data merging.
Ingestion points: The skill ingests content for text frames and external data for merging tasks via the data.source.select and data.merge operations as described in SKILL.md and references/control-protocol.md.
Boundary markers: There are no explicit instructions for the agent to use delimiters or specific warnings to ignore instructions that might be embedded within the document data or data sources.
Capability inventory: The skill has the capability to write local files (app.save, app.export), manipulate the application's user interface, and execute document-level commands via the control protocol (SKILL.md).
Sanitization: The instructions do not define specific sanitization or validation routines for data fetched from external sources before it is processed or merged into documents.
[COMMAND_EXECUTION]: The skill invokes local command-line tools and application binaries to manage sessions and perform document operations.
Evidence: The skill uses designcraft-cli and designcraft with specific arguments like --control and mcp --connect to establish automation sessions (SKILL.md).
[EXTERNAL_DOWNLOADS]: The skill documentation refers to and fetches configuration or documentation snapshots from a public GitHub repository.
Evidence: The skill mentions fetching the latest control protocol documentation from the storytold/designcraft repository on GitHub (SKILL.md).

designcraft