The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard utilities such as grep and ls to search through local project documentation and log files. These operations are restricted to the local file system and align with the skill's purpose of analyzing historical issues.
[DATA_EXPOSURE]: The workflow involves reading internal documentation and library source code (e.g., node_modules, site-packages) to diagnose root causes. This access is localized and necessary for developer-oriented debugging tasks.
[PROMPT_INJECTION]: The skill contains logic-enforcement instructions, such as the 'Circuit Breaker Rule' to prevent infinite loops of failed fix attempts. These instructions govern the debugging process and do not attempt to bypass the agent's core safety or operational constraints.
[SAFE]: No evidence of malicious intent, remote code execution from untrusted sources, or hidden data exfiltration was detected. The skill follows best practices by requiring explicit user confirmation before persisting any documentation or final fixes.

hard-fix