The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation and metadata contain imperative instructions directing the AI agent to follow specific operational protocols for task planning and user confirmation.
[PROMPT_INJECTION]: The skill processes external plan.md files from the local filesystem, creating an indirect prompt injection surface.
Ingestion points: plan-scanner.cjs and plan-parser.cjs read markdown files from user-defined directories.
Boundary markers: No specific delimiters or instructions are used to isolate content from these files in the data provided to the agent.
Capability inventory: The skill has the ability to execute system commands (execSync, spawn) and serve local files over HTTP.
Sanitization: HTML escaping is performed for the dashboard UI, but the raw file content is accessible via API and file-serving routes.
[COMMAND_EXECUTION]: The server script in server.cjs utilizes child_process.execSync to invoke system-specific browser commands and process.kill to manage server process lifecycles.
[REMOTE_CODE_EXECUTION]: The http-server.cjs script attempts to dynamically load modules using require with computed paths targeting a separate skill (markdown-novel-viewer), which introduces risks associated with runtime dependency on external directory structures.
[EXTERNAL_DOWNLOADS]: The skill specifies the gray-matter package as a required dependency to be installed from the npm registry.

plans-kanban