rc-skill-best-practices

# Writing Skills A skill exists to wrangle determinism out of a stochastic system. **Predictability** — the agent taking the same _process_ every run, not producing the same output — is the root virtue; every lever below serves it. **Bold terms** are defined in [`references/glossary.md`](references/glossary.md); look them up there for the full meaning. ## Branches | When you are… | Do | | --- | --- | | Creating a new skill from scratch | Read `references/authoring-procedure.md` in full first — spec mechanics, directory layout, metadata validation. Then apply the doctrine below to every line of content you write. | | Improving an existing skill (bloat, drift, weak triggers) | Apply the doctrine below, section by section, until every rule that bears on the skill has been applied. | | Diagnosing a skill whose references or assets the agent ignores | Read `references/loading-diagnosis.md` in full before proposing any fix — symptoms, the pointer-strength ladder, and the fixes in order. | Every branch ends the same way: a written audit, item by item. For create and improve, read `references/checklist.md` in full and mark each item Pass or Fail against the result; for diagnose, run the repair checklist in `references/loading-diagnosis.md`. Fix every Fail and re-audit — a skill that reads well but fails the audit is not done. The doctrine below is what Part A of that audit checks; it governs every line you write, not just the ones the procedure files mention. ## Invocation Two...