[caveman] Optimize instruction verbosity — agentic-chat, campaign, developer.instructions (2026-06-24)

[github-actions]

Summary

Editorial pass to reduce token verbosity across three instruction/agent prompt files. No rules, values, or behaviours were changed — only prose was tightened.

Why use many token when few do trick.

Changes

.github/agents/developer.instructions.md

• Removed redundant words and collapsed multi-line bullet lists into inline sentences throughout.
• Shortened checklist and section headings in the Prior Failure Analysis and Scope Hints sections.

.github/aw/agentic-chat.md

• Merged the "Stay Focused" item into the "Iterate" item, reducing the interaction guideline list from 6 to 5 items.
• Tightened phrasing in Core Principles, the description header, and the Final Step footer.

.github/aw/campaign.md

• Pruned redundant words across Stop Condition, Scope, Constraints, Pacing Levers, KPI Workflows, and Cross-repo Coordination sections.
• Collapsed multi-item bullets into inline prose where structure added no clarity.

Impact

• Functional impact: None — no logic, rules, values, or structured data were altered.
• Breaking changes: No.
• Test impact: None — prompt/instruction files only.
• Affected consumers: Any agent or workflow that reads these three files will see condensed instructions; meaning is preserved.

Generated by PR Description Updater for #41301 · 59.5 AIC · ⌖ 7.21 AIC · ⊞ 4.6K · ◷