The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes bash via the exec tool to perform a configuration check at the start of a session. It verifies the existence of the ~/.secrets/substack.env file and ensures that the required configuration keys are present using grep. This usage is restricted to setup verification.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from Substack posts, drafts, and comments, which represents an ingestion surface for potentially untrusted data.
Ingestion points: External data enters the agent context via tools such as substack-list_posts, substack-get_draft, and substack-get_post_comments (as referenced in SKILL.md).
Boundary markers: The instructions do not define specific delimiters or "ignore" instructions for the agent when processing retrieved content to mitigate potential command injection from Substack metadata or comments.
Capability inventory: The skill possesses capabilities to write to the local filesystem (during setup), execute shell commands, and create or publish content on Substack.
Sanitization: The instructions recommend converting ProseMirror JSON to markdown or plain text for presentation to the user, which provides a level of structural validation but does not explicitly include security sanitization for instructions.

substack-publish