The Agent Skills Directory

[SAFE]: The skill provides a structured methodology for logging decisions to a TSV file for audit purposes. The accompanying shell script scripts/log.sh includes specific protections against CSV injection (Formula Injection) by sanitizing cells that begin with formula-triggering characters (=, +, -, @) with a single quote. It also ensures the log file remains valid TSV by stripping newlines and tabs from input data. The instructions to read session transcripts are scoped to local project directories for self-audit purposes and do not involve data exfiltration or unauthorized access. The requirement for cross-model review is a safety feature intended to provide oversight on the agent's work.

show-me-your-work