Lessons and skill updates were being written with stiff, highly-specific content (exact function names, file paths, memory references) that narrows applicability and breaks self-containment. This tightens both skills to enforce abstraction.

usethis-lesson-create (v1.0 → v1.1)

• New "Self-containment and abstraction" section with three rules: never reference memories, never name specific code objects unless they are the subject, and state principles at the right abstraction level

• Comparison table showing "too specific" vs generalised rewrites:

  Too specific	Better
  Always call clear_functools_caches() after @functools.cache	Whenever a caching decorator is added, ensure the cache is cleared in tests
  validate_or_raise accepts err=, not error_cls	Validation helpers have specific argument conventions — check the signature before use

• Updated lesson template placeholder for "Generalised principle" to explicitly prohibit specific names and memory references

• Strengthened procedure step 3 with the same guidance

usethis-skills-modify (v1.5 → v1.6)

• New step 2 in the procedure: "Generalise the instruction" before writing skill content — if the incoming instruction names a specific function/file/class, lift it to a transferable principle first
• New "Generalising incoming instructions" section with a comparison table and the key test: "would this guidance still be relevant if the named entity were refactored away?"
• New "Never reference memories" content quality guideline — skill content must be fully self-contained