fix: accurate prompt-test claims, hook registration example and tighter README

[darthrootbeer]

Context

Second cold review of the repo found the top README misstating where prompt blocks exist and how they were tested, pattern prompts with no saved runs or good-answer lines, no settings example in the hooks pattern (a model guessed it wrong), and a dense first screen. Ticket: JBSR-099.

Changes

• Patterns: all nine prompts now have an A good answer ... sentence and were run on Sonnet and Haiku. Answers and grades are saved in patterns/tests/prompt-runs/, with the script that ran them. Each pattern's checked-how paragraph is now specific to that doc.
• Hooks pattern: adds the real settings.json registration example. The saved Haiku run now gets the shape right.
• Top README: accurate prompt-block coverage, three-sentence job-assessment entry, CI badge, no claim about line-by-line review.
• Job-assessment docs: sentence about diff review replaced with what the tests and CI check, quickstart uses a fresh output folder so it can be re-run, broken-profile output shown, merge-time claim dropped.
• Small fixes: CI actions bumped to current majors, typed-memory topic-file wording, plan-this wording, LICENSE name matches the commit author.
• Not touched: pipelines/docs-pipeline and skills/docs-readability-check.

🤖 Generated with Claude Code

https://claude.ai/code/session_01XV3Ggh86cf2xwUyz28XaN6

• Docs updated