docs(research): add 2026-05-26 GPT-5.5 Codex audit of Skill System

Parallel SYSTEM-mode audit by GPT-5.5 in Codex, run in parallel with
Opus 4.7 (workspace docs/reports/skill-system-audit-2026-05-26.md).
Posted as comment on SH-6525.

18 findings (F1-F18). Strongest finding Opus missed: check-audit-manifest.js
fails with 16 historical graded-comprehension verdicts without backing
artifacts, AND it's outside npm run verify — filed as SH-6548.

Synthesis at workspace docs/reports/skill-system-audit-synthesis-2026-05-26.md
merges both audits into one combined view (23 distinct underlying issues,
9 Codex-unique, 5 Opus-unique, 9 overlap).

Refs SH-6525, SH-6539

Co-Authored-By: GPT-5.5 (Codex) <noreply@openai.com>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: 3 people