The Agent Skills Directory

[PERSISTENCE_MECHANISMS]: The skill instructs the user to execute a bootstrap script that modifies the agent's settings.json file (either globally in ~/.claude/ or at the project level). It installs hooks for SessionStart and UserPromptSubmit that automatically execute the hooks/inject-writing-standards.py script. This establishes a persistent execution channel within the agent's core operational loop.
[DATA_EXPOSURE_AND_EXFILTRATION]: The bootstrap process reads and modifies the agent's platform configuration file (settings.json). This is a sensitive system component that governs the agent's behavior, hooks, and environment settings.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and review external documents (READMEs, specs, etc.) to enforce writing standards. This process lacks boundary markers and sanitization, creating a surface where instructions embedded in untrusted prose could influence agent behavior.
Ingestion points: Document content provided for review or editing tasks.
Boundary markers: None identified in the instructional rules or hook scripts for separating untrusted content from system instructions.
Capability inventory: The skill has established the ability to modify system settings and execute scripts automatically on every prompt via the bootstrap process.
Sanitization: No sanitization or filtering of external document content is performed prior to processing.
[COMMAND_EXECUTION]: The SKILL.md file directs the user to perform manual command execution of the bootstrap.py script to wire the automated reminders.

writing-standards