The Agent Skills Directory

General Security Posture (SAFE): The skill is a well-structured instructional methodology based on a trusted model context protocol server. It focuses on enhancing the agent's internal reasoning process without requiring elevated system privileges.
Indirect Prompt Injection (INFO): While the skill ingests 'thoughts' that could theoretically contain data from untrusted sources, the processing scripts (process-thought.js and format-thought.js) are limited to validation and string formatting. They do not possess sinks for remote code execution or data exfiltration.
Ingestion points: CLI arguments in scripts/process-thought.js.
Boundary markers: Relies on agent-provided thought headers (e.g., 'Thought 1/5').
Capability inventory: Local file writing (.thought-history.json) and console logging only.
Sanitization: Basic data type and string presence validation.
Data Exposure (SAFE): Thought history is maintained in a local file within the script directory. There are no patterns indicating sensitive file access or transmission of data to external domains.
Dependency Analysis (SAFE): The project includes jest as a development dependency for unit testing. No production dependencies or remote script execution patterns were detected.

sequential-thinking