[caveman] Optimize instruction verbosity — syntax-tools-imports, test-coverage, token-optimization, triggers (2026-06-22)

[github-actions]

Summary

Prose-only condensation pass across four .github/aw/ reference documents. No technical content, instructions, or techniques were added, removed, or altered — every change is a wording tightening.

Changes

File	What changed
.github/aw/syntax-tools-imports.md	Tightened sentences throughout; collapsed bullet lists; removed redundant explanations.
.github/aw/test-coverage.md	Removed bold emphasis from the "always prefer pre-computed artifacts" principle; now plain prose.
.github/aw/token-optimization.md	Substantially shortened every major section (frontier-model guidance, DataOps, sub-agent, caveman technique, batching, observability). No techniques dropped or redefined.
.github/aw/triggers.md	Condensed fuzzy-scheduling prose, weekday-schedule rationale, and the issue_comment/pull_request_comment mapping note; multi-bullet lists collapsed to single sentences.

Impact

• Breaking changes: None
• Behaviour changes: None — documentation only
• Token impact: Net reduction; this is a direct application of the caveman technique documented in token-optimization.md itself

Testing

No functional code was changed. A quick read-through of each file is sufficient to verify intent was preserved.

Generated by PR Description Updater for issue #40871 · 70.2 AIC · ⌖ 6.58 AIC · ⊞ 4.5K · ◷