The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to read and analyze user-generated content across various story levels (e.g., synopsis, outline, manuscript). This content is treated as data to detect inconsistencies, which creates an indirect prompt injection vector where an attacker could place instructions within the story text to manipulate the agent's review or its updates to the project dashboard.\n
Ingestion points: The agent is instructed to read files in the pitch/, structure/, scenes/, entities/, and manuscript/ directories, as well as the change-log.jsonl file.\n
Boundary markers: The skill does not implement delimiters or explicit instructions for the agent to ignore potentially malicious commands embedded in the story data.\n
Capability inventory: The skill uses local Deno scripts to perform file system operations (read, write, directory creation, and file watching) and narrative analysis tools.\n
Sanitization: No sanitization or filtering of the story content is performed before processing.\n- [COMMAND_EXECUTION]: The skill provides utility scripts (init.ts, watcher.ts, status.ts) and guides the user to execute them via the Deno runtime. These scripts perform necessary file management tasks, such as initializing directory structures and logging markdown file modifications, which are appropriate for a story synchronization tool.

story-zoom