Use after completing any substantive piece of work (code change, analysis, document, configuration, answer to a hard question) and BEFORE presenting it as done. Switches you from author to attacker - you try to refute your own work and only present it if it survives. Do not skip because the work "looks clean"…
Use whenever you are about to assert or act on a fact about a system - what a function does, what a config is set to, what is deployed, what a table contains, what a dependency's version is. Forces verification against the LIVE system before the assertion. Docs, comments, READMEs, and your own memory are treated as…
Use when writing to or reading from any persistent memory an agent keeps across sessions - CLAUDE.md, memory files, notes, project docs meant for future sessions. Governs WHAT deserves persisting, HOW to write it so it survives time, and the recall rule: verify remembered facts against live state before acting on…
Use before starting ANY task that involves more than one edit, file, or step - coding, refactoring, writing, configuration, migrations. Forces a short written plan (goal, unknowns, success criteria, step order) BEFORE the first change. Do not use for single-line fixes or pure questions.
Use whenever you produce prose a human will read - documentation, READMEs, reports, summaries, commit messages, emails, blog drafts, PR descriptions. Runs a cutting pass where every sentence must earn its place. Target: 30 percent shorter with zero information loss. Do not use on code or on text the user asked you to…
Use on every task where you will modify existing work - code, documents, configs. Fences your changes to exactly what was asked. Adjacent problems get FLAGGED, never silently fixed. Keeps diffs minimal and reviewable. Do not use it to refuse legitimate follow-through the task actually requires.
★not rated 24 2mo agoA64 tokens
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: