[caveman] Optimize instruction verbosity — charts.md (2026-05-07)

[github-actions]

Caveman Optimization Run — 2026-05-07

Applies the caveman optimization principle to instruction files:

"Why use many token when few do trick."

Files Optimized

File	Before	After	Removed
.github/aw/charts.md	~663 lines	~592 lines	Redundant prose throughout

What Was Changed

Seven targeted cuts from .github/aw/charts.md:

1. Opening paragraph removed — restated the frontmatter description field verbatim.
2. Option B library/directory intro condensed — replaced 19-line verbose section (library list + directory tree code block) with 2-line compact format already used in Option A.
3. Chart Quality Guidelines bullets removed — the preceding code block already demonstrates every setting (dpi=300, seaborn whitegrid, etc.).
4. "You are an expert..." preamble removed from Option C agent instructions — vacuous identity statement, adds no schema signal.
5. Error Handling section removed from Option B — generic Python file/column validation, not AW-specific schema.
6. Trends Visualization Best Practices bullet subsections removed — three subsections stated obvious charting conventions already shown by the code examples that follow.
7. Tips for Success section removed — all 7 tips restate rules already present earlier in the file (ISO 8601 timestamps, 90-day retention, data-first pattern, DPI 300, etc.).

What Was Preserved

All frontmatter examples, YAML tool configuration, code blocks showing AW schema, safe-output patterns, the upload-asset workflow, cache-memory patterns, session analysis chart specs, and all compiler constraints were kept intact.

Round-Robin Progress

Processed files 5–9 of 26 total files in the queue (indices 4–8). Next run starts at index 9 (.github/aw/experiments.md).

🤖 Generated with [Claude Code]((claude.ai/redacted)

Generated by Daily Caveman Optimizer · ● 570.9K · ◷

• expires on May 10, 2026, 9:16 PM UTC