The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs emphatic instructional language, such as "MANDATORY IMPORTANT MUST ATTENTION," to prioritize specific workflows like task planning and citing evidence. While these patterns use high-attention keywords, they are used to enforce quality control and operational rigor rather than to bypass safety protocols or extract sensitive system information.
[DATA_EXFILTRATION]: The skill instructions involve reading codebase changes and writing journal entries to the local ./docs/journals/ directory. No network-capable commands, external URLs, or patterns indicating data exfiltration were found.
[COMMAND_EXECUTION]: The workflow leverages structured tools like TaskCreate for project management and a specialized journal-writer subagent. It does not invoke shell commands, perform system-level configurations, or utilize risky subprocess patterns.
[SAFE]: No evidence of obfuscation, hardcoded credentials, remote code execution, or persistence mechanisms was detected in the skill configuration or instructions.

journal