The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a shell script scripts/init-workspace.sh used to initialize the local coaching environment.
The script creates the thinking-gym/ directory structure and populates it with markdown templates for the user's profile and calibration logs.
The operations are idempotent and use standard shell utilities without network access or elevated privileges.
[PROMPT_INJECTION]: The skill processes untrusted documents provided by the user for auditing, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted text is ingested during the 'Intake' step described in SKILL.md when the user provides a document for auditing.
Boundary markers: The instructions do not define explicit delimiters or isolation markers to encapsulate the audited document content.
Capability inventory: The skill records audit results, updates a persistent reasoning profile (profile.md), and maintains a calibration log (calibration.md) in the local filesystem.
Sanitization: The instructions mandate strict 'quote discipline' (character-for-character quoting) to ensure findings are anchored to literal text, which serves as a logical verification step during the manual auditing process.

critical-thinking