The Agent Skills Directory

[SAFE]: The skill implements security-conscious design patterns, including a human-in-the-loop confirmation step (Step 3) before executing any work, and a preflight script that ensures the agent has the necessary permissions without attempting to self-grant them.
[COMMAND_EXECUTION]: The skill executes local scripts (scripts/preflight.sh) and system utilities (git, jq, work-item-tracker.sh) to perform its primary functions. These commands are well-scoped and essential for project task management.
[DATA_EXFILTRATION]: No evidence of unauthorized data harvesting or network exfiltration was found. The preflight script's access to configuration files (settings.json) is limited to verifying permission states for the preflight report.
[PROMPT_INJECTION]: The skill instructions contain no patterns of prompt injection, jailbreak attempts, or safety guideline overrides. It follows a structured, task-oriented logic.
[EXTERNAL_DOWNLOADS]: The skill does not download or execute remote code from external sources. It relies on local scripts and pre-installed system tools.
[REMOTE_CODE_EXECUTION]: There are no identified patterns for dynamic code generation or remote command execution. The execution logic is static and utilizes known local paths.

work