The Agent Skills Directory

[SAFE]: The skill follows security best practices for managing agent instructions, including explicit validation steps and strict formatting requirements.
[COMMAND_EXECUTION]: The skill performs targeted file system operations to create, edit, or remove rule files within project directories (specifically .claude/rules/, .cursor/rules/, and AGENTS.md). These operations are constrained by an 'itemized' editing policy that prevents broad, destructive writes and ensures that only one rule is modified at a time.
[PROMPT_INJECTION]: The skill manages the generation of agent-guiding instructions. To mitigate the risk of improper or redundant instruction injection, it enforces an 'Admission Filter' that rejects drafts that are not 'non-discoverable' or 'mistake-correcting'. It also imposes a line-count budget to prevent complex, overreaching instructions from degrading agent performance or safety.

rule-writing