The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute an external binary (anywrite) provided by the user. It explicitly requires an absolute path and includes checks to prevent the execution of potentially malicious artifacts from within the project directory (e.g., ./dist/anywrite).
[CREDENTIALS_UNSAFE]: The skill instructions reference the storage of an API key in ~/.anywrite/config.json and the use of the ANYTYPE_API_KEY environment variable. While these are standard practices for CLI tools, the agent has the capability to interact with these sensitive local files.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests untrusted data from an external source (Anytype objects, chat messages, and search results).
Ingestion points: Content is retrieved from the Anytype local API via the objects, chat, and search resources.
Boundary markers: The instructions do not specify the use of delimiters or specific instructions to ignore embedded commands in the retrieved data.
Capability inventory: The skill can execute the anywrite CLI, write to Anytype objects, and upload files.
Sanitization: There is no mention of sanitization or filtering of the content retrieved from the Anytype API before it is processed by the agent.

anywrite