The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests repository data for architectural analysis. Ingestion points include files like package.json, Makefile, CI config, Dockerfile, and deployment manifests. No specific boundary markers or sanitization logic for this external content are defined. The skill provides analysis and can delegate implementation tasks to the $tk agent.
[DATA_EXFILTRATION]: The skill reads potentially sensitive information from configuration files and accesses session history logs in ~/.codex/sessions for auditing and maintenance.
[COMMAND_EXECUTION]: The maintenance procedures reference a custom utility, seq, to query session history and generate reports based on specific user triggering phrases such as 'over-engineered' or 'start simpler'.

reduce