The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a helper script scripts/log.sh to manage the decision log. This script follows shell scripting best practices, including set -euo pipefail and input cleaning to maintain log integrity.- [INDIRECT_PROMPT_INJECTION]: The logging script includes a dedicated cleaning function designed to prevent CSV/TSV formula injection attacks. It sanitizes inputs by stripping control characters and prefixing potential formula starters (like =, +, -, and @) with a single quote, protecting users who view the resulting log in spreadsheet applications.- [DATA_EXPOSURE]: The skill includes instructions for the agent to audit its work by reading transcript files located in ~/.claude/projects/. This access is specifically limited to the current project's directory to verify the truthfulness of the decision log, and it contains explicit warnings against accessing unrelated project data.

show-me-your-work