docs(wiki): Claim 11 — multi-signal language detection (#3119)

[squid-protocol]

Closes #3119.

What

Writes up the multi-signal detection cascade in language_lens.py as Claim 11 in the 03- series — the argued-position job that 02-05-language-lens.md (mechanical reference) deliberately doesn't do. New page: docs/wiki/03-11-claim-11-multi-signal-detection.md.

Follows Claim 10's intellectual-honesty structure:

1. Thesis — a file extension is an assertion, not evidence; detection is inference with an explicit trust level, not a waterfall returning a bare label.
2. The signals combined — all zero-dependency (exact-filename, extension w/ wrapper unwrapping, shebang, prose anchors, sibling anchors, ecosystem gravity, internal_discriminator, lexical scan, discovery funnel, context priors).
3. The three genuinely differentiated properties — agreement-based tiering with lock_tier + source_proof provenance; identity-conflict-as-refusal (Tier 5 + Identity Masking security flag); repo-context inference (incl. the Ecosystem gravity should weigh sibling classifications, not extension counts #3137 sibling-classification vote — an EQU-only copybook is HLASM because of where it sits).
4. Where it is convergent, not novel — states plainly that the collision-registry mechanism is functionally Linguist's heuristics.yml and the lexical self-scoring is Pygments' analyse_text(); frames independent arrival as validation, not invention.
5. Honest limits — the hand-tuned, uncalibrated lexical constants (+= 15.0 on a substring test, two 1.25× boosts, the 0.4× abap/fortran/cobol handicap, log1p normalisation, /50 confidence — all verified in source), plus the four harness-found defects (Detection hardening: EQU-only .asm copybooks misclassify as assembly (hlasm discriminator gap) #3110, Shebang matching is unanchored substring search: canonical tclsh/wish/ts-node shebangs trigger false Identity Conflicts (Tier 5) #3116, Pin internal_discriminator and shebang resolution order for contested extensions #3118, Ecosystem gravity cannot resolve a same-extension collision: the contested extension votes for one of its own claimants #3132) the engine has since fixed.
6. The number — §6 of Write up the multi-signal language-detection architecture as Claim 11 #3119 soft-blocked this on a measured figure (Add a language-detection accuracy harness (per-language precision/recall + confusion matrix + committed baseline) #3117). That harness now exists and, after Ecosystem gravity cannot resolve a same-extension collision: the contested extension votes for one of its own claimants #3132/Ecosystem gravity should weigh sibling classifications, not extension counts #3137, reports 1.0 determinable (3,231/3,231), 509/509 on the hand-labelled contested subset, and an empty confusion matrix — with the extension-circularity and "generalisation estimate, not universal proof" caveats stated inline.

Wiring

• mkdocs.yml nav (between Claim 10 and Future Outlooks)
• docs/wiki/index.md — both the Systems-Engineers list and the numbered claims list
• docs/document_alignment_guide.md — new claim-map row + the "10 → 11 numbered Claims" count
• Reciprocal cross-link with 02-05-language-lens.md (and fixed that page's broken back-to-index footer)

Validation

mkdocs build --strict shows zero warnings on any of the touched/added files; the new page and all its links resolve. (The build's 8 pre-existing warnings are unrelated dead ../vectors.md / ../zero_dependency_mode.md links in other pages, present before this PR.)

Docs only — no engine change.

🤖 Generated with Claude Code

[github-actions]

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab