[davehallmon]

What this changes

Prepares the 1.2.0 release. Docs only; SKILL.md is the same as in #10.

• CHANGELOG: Unreleased becomes [1.2.0] — 2026-10-04. Tested With now has two rows for 1.2.0:
  • Claude Opus 5.5, eval rounds 4–7.
  • Claude Sonnet 5.5 on claude.ai on the web, the manual check.
• README:
  • Version badge 1.2.0.
  • Results table through round 7.
  • A plain summary: about even with plain Claude overall, leads on honesty and actionability, trails on fit, diagnosis, and reading time.
  • The two rules tested in 1.2.0, and a line on the claude.ai check.
• index.md: current results line.
• evals/manual/: a record of the claude.ai check and the full Test A response.

The claude.ai check (Sonnet 5.5, web)

• Full run: pass.
• Decision prompt: rendered as clickable multi-select choices.
• Learn more: all six links come from the curated file. Four were opened and loaded.
• Fact question: no Skill. Trivial problem: the Skill loaded on its own, without the slash command, and declined the table.
• Caveat: Test A itself was started with /problem-lens, so automatic triggering is shown by the printer test, not by Test A. The record says so.

Order

1. Merge Skill+evals: number guard and round 7 #10 (number guard + round 7).
2. Retarget this PR to main and merge.
3. Then the v1.2.0 tag and GitHub release go up with problem-lens.zip attached.

🤖 Generated with Claude Code

https://claude.ai/code/session_01X9xscdGXL5xQmmxnwEf9T7

---

Generated by Claude Code