SKT-0060: Read dump metric names against documented signal formats with explicit unresolved evidence

Description

The scratch conformance reader leaves 2330 names unresolved because it cannot parse prose declarations, dynamic expansions and native OTLP documentation. That uncertainty must be measured separately from proven synth defects.

Acceptance Criteria

#1 A supported command and tested reader handle documented YAML, prose and expansion formats without changing emitters; root-owned syntax and sourced catalogue inventory corrections retain exact telemetry names.
#2 An explicit-selection dump measurement reports resolved and unresolved names and every named parse gap, with limits on label and envelope claims.
#3 Focused tests, the integrated gate and CodeRabbit review complete before source integration.

Definition of Done

#1 just check (fmt-check, lint, gen-check, env-check, docs-check, test, race, hygiene, ui-check, compose-check, helm-test, lab-check, signal-fidelity)
#2 just gen (only if a blueprint field, construct/workload config struct, or a skill under plugins/synthkit/skills/ changed)
#3 just dump — inventory diffed against signals/

Implementation Plan

Promote the reader into a tested package, pin format cases with fail-first tests, measure actual dump output, wire the just entrypoint and review the bounded batch.

Parse every fenced yaml signals block with YAML v3, including multi-document blocks, recoverable malformed mappings, CloudWatch stats/info expansions, and native OTLP declarations.
Parse prose tables and backtick/brace alternatives into transport-specific documented names while recording each named unsupported format as a parse gap.
Parse the stable -dump sections with sorted label or attribute keys, compare names by Prometheus and OTLP transport, and expose resolved, unresolved, and gap evidence with explicit scope limits.
Add a small CLI under internal/conformance and a just recipe that reads signals/ and stdin without mutating emitters or catalogue files.
Run focused tests, the explicit-selection dump measurement, and the proportionate repository checks; review the final diff before task finalization.

Wave 2026-09-14: diagnose eight parse gaps, apply syntax-only catalogue fixes at the root wiring pass, then remeasure the retained explicit selection before evaluating Envoy and CloudWatch name-resolution hypotheses. Keep name evidence separate from label and envelope conformance; classify every remaining unresolved name with its evidence boundary.

Implementation Notes

Final integration selection adds the P3 reference blueprint: 2973 resolved of3347 names,374 unresolved and8parse gaps. The earlier2972/3346 observation remains the measurement before that selection addition.

Eight syntax gaps closed without telemetry-name changes; syntax alone leaves 374 unresolved. Reader resolves 3269/3347; current normative Envoy inventory yields 3317/3347 with 30 remaining. Fresh P3-inclusive dump: 3327/3357, 154 blocks, zero gaps, same 30 classified names. CloudWatch mapping resolves 122/126 source names; Envoy parse-gap hypothesis disproved. No emitter changed for reader findings. Fail-first tests, CodeRabbit disposition and full gate retained in codex/scratch/wave-2026-09-14/. Name resolution only; label/envelope conformance unproven. AC wording aligns with goal-authorized syntax and sourced catalogue corrections.

Wave catalogue correction: all 103 already-sourced dotless Azure native names now have twelve additive sink: otlp blocks (119 rows, 103 unique roots, no stats), preserving the tables and reader. Retained September-13 selection is 3347/3347; fresh preflight wildcard selection was 3606/3709 with 103 unresolved; integrated fresh wildcard selection is 3711/3711 with no parse gaps. Additional observed-only managed names do not change that dump result. This supersedes older unresolved-count summaries without rewriting their historical measurements. Name resolution does not validate envelopes. Source integration bcf1687689fc9199e8f232b3cef23370b56f5aba; exact-source CI34279467976 success. Evidence: codex/scratch/wave-2026-09-16/conformance-fixed-integrated.json, conformance-live-integrated.json and l1/brief.md.

Final Summary

Integrated in 4dbe491. Reader command: just signals-conformance dump.txt. Explicit-selection measurement resolves2972 of3346 metric names, leaving374 unresolved and8named YAML parse gaps. Name evidence only; dumped label keys are retained but not compared, and unresolved names are not proven emitter defects. Focused tests and isolated integrated just check passed. CodeRabbit minor indentation finding fixed; no remaining finding. No blueprint schema change was required by the reader.

Current catalogue evidence: retained selection3347/3347 and integrated fresh wildcard3711/3711, no unresolved names or parse gaps. Reader source unchanged; additive sourced Azure native blocks landed in bcf1687689fc9199e8f232b3cef23370b56f5aba.