The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard shell utilities (jq, mkdir, git, rm) for configuration management and concurrency locking. These are used in a controlled manner within the repository's .git directory to prevent race conditions during automated loops.
[REMOTE_CODE_EXECUTION]: The skill uses a templates/resolve-config.sh script to merge configuration files. This script is localized to the repository, uses the standard jq utility, and contains no remote fetch or execution logic.
[DATA_EXFILTRATION]: The skill interacts with GitHub, GitLab, and Linear via standard CLI tools (gh, glab) or MCP servers. These interactions are limited to the intended purpose of reading and updating issue statuses and PR reviews. No sensitive data access outside of the project scope was detected.
[PROMPT_INJECTION]: The SKILL.md contains an 'Author authority' section that explicitly instructs the agent to treat third-party text (comments, issue bodies) as data rather than instructions. This is a defensive measure against indirect prompt injection.
[CREDENTIALS_SAFE]: The skill manages configuration via .tituskirch-skills.json and uses .env files for secret management as referenced in the guidelines, which is a standard and safe practice.

work-review-queue