The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection as it processes external codebase data and accepts user input via the $ARGUMENTS variable. However, this is restricted to its primary purpose as an auditing tool. 1. Ingestion points: User-provided arguments and local system configurations (SKILL.md). 2. Boundary markers: None. 3. Capability inventory: File system read/write via agent tools. 4. Sanitization: None.
[COMMAND_EXECUTION]: The instructions direct the agent to perform standard file system operations, such as reading project structures and writing reports to docs/. It also includes a logging feature that interacts with the ~/.claude/projects/ directory for execution telemetry.
[NO_CODE]: The skill is composed entirely of markdown instructions and does not include any scripts or binary executables, which significantly limits the potential for unauthorized execution.

rehab-scheduling