[ambient-context] Daily Ambient Context Optimizer - 2026-07-09

[github-actions]

Executive Summary

• 4 runs sampled across 4 distinct workflows (July 9, 2026)
• Median first-request size: 13,467 chars; P95: 19,183 chars
• System boilerplate dominates: 49–70% of every prompt is framework-injected (<safe-outputs> 3,481 chars, <mcp-clis> ~1,083 chars) — not reducible via workflow .md changes
• Highest-cost workflow: Daily Code Metrics (AIC=236.84, 19,183 chars, 28 headings, 4.3% dup-line ratio)
• Top opportunity: condense verbose prose sections in daily-code-metrics.md for est. ~1,800 chars savings

Highest-Leverage Changes

1. Condense daily-code-metrics.md verbose data sections — ## Metrics to Collect (984 chars), ## Data Storage (789 chars), and # Prune history.jsonl (702 chars) can be rewritten as compact tables/bullets. Est. reduction: ~1,800 chars (~9% of compiled prompt).
2. Trim shared/trends.md chart-style overlap — 1,530 chars compiled into Daily Code Metrics; prose duplicates shared/python-dataviz.md. Reduce to chart-type list only. Est. reduction: ~700–900 chars (within 40% guard).
3. Compress pr-code-quality-reviewer.md step prose — Steps 1–5 total 5,138 chars; Steps 1 and 4 alone are 2,880 chars of multi-paragraph prose that can be condensed to tight bullets. Est. reduction: ~900–1,200 chars.
4. Move Python path examples to pre-agent steps in daily-code-metrics.md — ### Python Script (529 chars) is partially redundant with the existing Write chart specs pre-agent step. Est. reduction: ~400 chars.
5. Verify no double-import of shared/trends.md — check whether shared/trends.md and shared/python-dataviz.md both compile into the same prompt (low risk, verify with make recompile diff).

CI-Validation Checklist for Implementing Agents

Any agent implementing workflow-file recommendations must complete every item below before opening a PR:

• Run make recompile for every modified .github/workflows/*.md file — zero compilation errors required
• Confirm no target workflow file is reduced by more than 40% of its current character count
• Run make agent-report-progress before the final commit and confirm it passes
• Check the blocked_files list in /tmp/gh-aw/ambient-context/closed-pr-targets.json (written by Step 4) — do not re-attempt changes to any file that appears in a closed ambient-context optimization PR from the last 14 days
• Include a smoke-test result or dry-run output in the PR description for every changed workflow spec
• Document all generated .lock.yml changes in the PR body
• Do not submit as a draft PR

Key Metrics

Metric	Value
Sampled runs	4
Distinct workflows	4
Median chars	13,467
P95 chars	19,183
Largest sampled request	19,183 chars (Daily Code Metrics, AIC=236.84)
Merged optimizer PRs (7d)	3
Closed optimizer PRs (7d)	0
Optimizer PR close-rate (7d)	0%

Per-Run First-Request Metrics

Run	Workflow	Chars	Headings	Code Fences	Dup Line%	AIC	System%
§29042654541	Daily Code Metrics	19,183	28	12	4.3%	236.84	49.6%
§29039556698	PR Code Quality Reviewer	14,013	16	8	2.5%	227.06	49.1%
§29043227257	Daily Safe Output Integrator	12,589	15	4	1.5%	96.81	66.7%
§29040714758	Daily Testify Uber Super Expert	12,921	19	4	1.3%	92.07	70.0%

Note: API proxy event logs not present; prompt.txt used as source.

Repeated Ambient Context Signals

• <safe-outputs> block: 3,481 chars in all 4 runs — framework-managed
• <mcp-clis> block: ~1,083 chars in all 4 runs — framework-managed
• 31 lines >30 chars common to all 4 runs (~6,171 chars) — all framework-injected
• shared/reporting.md (497 chars) included in 3/4 runs — already well-compressed
• shared/trends.md (1,530 compiled chars) overlaps with python-dataviz.md on chart style prose

Deterministic Analysis Output

Script: /tmp/gh-aw/ambient-context/analyze_requests.py (stdlib only)

• Highest dup-line ratio: 4.3% (Daily Code Metrics) vs 1.3–2.5% for others
• System block share: 49–70% of compiled prompt (all framework-injected, not reducible here)
• Daily Code Metrics: 28 headings, 12 code fences — most structural complexity of any sampled run
• PR Code Quality Reviewer: 1 inline agent (1,111 chars), 5 step sections (5,138 chars total)

Recommendations by Category

Workflow Markdown

1. Condense daily-code-metrics.md data sections (high, needs_review)

• File: .github/workflows/daily-code-metrics.md (11,046 chars; max reduction: 4,418 chars)
• ## Metrics to Collect + ## Data Storage + # Prune history.jsonl = 2,475 chars of prose → ~400 chars as compact tables
• Est. reduction: ~1,800 chars from compiled prompt

2. Trim shared/trends.md overlap (medium, needs_review)

• File: .github/workflows/shared/trends.md (1,992 chars; max reduction: 797 chars)
• Chart-style guidance (figsize, palettes, annotations) duplicates shared/python-dataviz.md
• Keep repo-specific chart type list; drop generic styling prose
• Est. reduction: ~700 chars from Daily Code Metrics prompt

3. Compress pr-code-quality-reviewer.md step prose (medium, safe)

• File: .github/workflows/pr-code-quality-reviewer.md (10,116 chars; max reduction: 4,046 chars)
• Steps 1 and 4 total 2,880 chars as prose paragraphs; rewrite as numbered bullets
• Est. reduction: ~900–1,200 chars

4. Move Python path examples to pre-agent step (low, safe)

• File: .github/workflows/daily-code-metrics.md
• ### Python Script (529 chars) overlaps with the existing Write chart specs pre-agent step
• Consolidate into pre-agent step script; remove from prompt body
• Est. reduction: ~400 chars

Skills

5. Verify import deduplication for python-dataviz.md (low, safe)

• shared/trends.md may re-import python-dataviz.md creating double-compilation
• Run make recompile and diff compiled output before/after; no workflow text change needed

References

• §29042654541 — Daily Code Metrics (AIC=236.84)
• §29039556698 — PR Code Quality Reviewer (AIC=227.06)
• §29043227257 — Daily Safe Output Integrator (AIC=96.81)

Generated by 🌫️ Daily Ambient Context Optimizer · 162.8 AIC · ⌖ 24 AIC · ⊞ 7.7K · ◷

• expires on Jul 16, 2026, 12:34 PM UTC-08:00

[github-actions]

This issue is being closed as outdated. A newer issue has been created: #44834

View newer issue

---

This action was performed automatically by the Daily Ambient Context Optimizer workflow.