[cschanhniem]

Implements F6 and A7 from NEXT_100_PR_MAP.

What

• docs/evidence/NEGATIVE_RESULT_EXAMPLES.md (F6): 5 worked examples contrasting informative vs non-informative negative results; hemolysis rejection (informative), module timeout (non-informative), dual-use flag (informative+mandatory), ad-hoc removal (non-informative), low-confidence borderline; summary table; agent behavior rules
• Makefile (A7): make docs-check target combining doc-links-check and docs-index-check; one-command docs validation for agents

Why

F6: Without concrete examples, users can't reliably distinguish when a negative result is worth recording vs when it's a software error. The distinction matters because non-informative results inflate rejection rates and corrupt the calibration loop.
A7: Without make docs-check, agents must remember to run doc-links-check and docs-index-check separately. A single target reduces missed checks.

Disconfirming pass

• F6 examples do not add real candidate data (all use TOY- prefix)
• A7 target does not change any safety gate or threshold — it only combines two existing checks
• Example 2 and 4 explicitly mark rejections as non-informative rather than pretending they are