The Agent Skills Directory

[PERSISTENCE]: The skill modifies the user's home directory configuration (~/.codex/hooks.json) to register persistent event hooks. This allows the skill's Python scripts to be executed automatically by the platform on session events such as SessionStart and Stop to facilitate communication with the local UI server.- [EXTERNAL_DOWNLOADS]: The instructions recommend and execute a remote shell script from a well-known service (https://astral.sh/uv/install.sh | sh) to install the 'uv' tool required for managing the skill's dependencies.- [COMMAND_EXECUTION]: The skill frequently executes local CLI commands via uv run to manage the lifecycle of the local reading server, process document content, and manage session hooks.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied Markdown and text documents which may contain embedded instructions designed to influence the agent's behavior during tutoring sessions.
Ingestion points: Document content is read from local paths specified by the user in tools/src/active_reader/store.py.
Boundary markers: The skill instructions do not specify markers to delimit document content from system instructions when processing text.
Capability inventory: The skill can execute subprocesses, perform file system operations, and make local network requests.
Sanitization: The server-side rendering logic in tools/src/active_reader/server.py disables HTML parsing in Markdown to prevent XSS within the reader UI.

reread