The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses emphatic instructional language (e.g., 'MANDATORY IMPORTANT MUST ATTENTION') to enforce strict adherence to its multi-phase workflow. These are standard prompt engineering techniques used to maintain consistency in complex operations and do not represent attempts to bypass safety filters.
[COMMAND_EXECUTION]: The skill includes instructions to execute a local Python script (python .claude/scripts/code_graph) for tracing code dependencies. This is a legitimate use of internal project tooling for static analysis.
[DATA_EXFILTRATION]: The skill reads local project documentation and configuration files (e.g., docs/project-config.json) to understand the codebase context. No network operations or external data transmission patterns were found.
[SAFE]: The skill implements a 'Self-Review Gate' requiring the agent to self-invoke a code review skill on any files it modifies. This is a security best practice that ensures changes are verified before completion.

code-simplifier