The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local Node.js scripts (scripts/process-thought.js and scripts/format-thought.js) to validate, track, and format thought sequences. These scripts are executed by the agent to maintain state and provide structured output.
[INDIRECT_PROMPT_INJECTION]: The skill presents an ingestion surface for potentially untrusted data through the "thought" content processed by the scripts.
Ingestion points: The --thought command-line argument used when invoking the thought processing scripts.
Boundary markers: Absent. The scripts treat input thought text as raw strings without implementing specific delimiters or "ignore" instructions.
Capability inventory: The skill possesses local file-write capabilities (writing to .thought-history.json) and formatted console output. It does not have network access, general system command execution, or elevated privileges.
Sanitization: Absent. The scripts perform basic data type validation but do not sanitize or filter the content of the thoughts for malicious instructional patterns.
[SAFE]: The skill uses a local file (scripts/.thought-history.json) for data persistence across thoughts within a session, which is a standard functional requirement for the described methodology. No network exfiltration or sensitive file access was detected.

sequential-thinking