Extract inline output templates from daily-cache-strategy-analyzer into shared file

[Copilot]

The Phase 5 (issue body) and Phase 6 (discussion body) template blocks were unconditionally loaded into every run's initial prompt, inflating first-request context by ~3.5 KB even when no findings existed.

Changes

• New: .github/aw/cache-strategy-templates.md — shared file containing the full issue and discussion body templates, organized under ## Issue Template and ## Discussion Template sections
• Updated: daily-cache-strategy-analyzer.md — replaced both inline template blocks with single-line read-on-demand references; the agent reads the templates file only when it is about to create output

<!-- Phase 5 — before -->
### Issue Template
**Title**: `[cache-strategy] Fix cache miss in <workflow-name>`
**Body**:
```markdown
### Cache Strategy Problem: <workflow-name>
... ~50 lines ...

Read the issue template from .github/aw/cache-strategy-templates.md
(see "## Issue Template" section) immediately before creating each issue.

**Size**: 15,337 → 12,462 chars (−18.7%), above the 40% reduction guard. Lock file recompiled.

<!-- START COPILOT CODING AGENT SUFFIX -->

- Fixes #49846

---

Run: https://github.com/github/gh-aw/actions/runs/30772836840

> Generated by [👨‍🍳 PR Sous Chef](https://github.com/github/gh-aw/actions/runs/30772836840) · gpt54 · 16.7 AIC · ⌖ 6.74 AIC · ⊞ 8.3K · [◷](https://github.com/search?q=repo%3Agithub%2Fgh-aw+%22gh-aw-workflow-call-id%3A+github%2Fgh-aw%2Fpr-sous-chef%22&type=pullrequests)
> <sub>Comment <em>/souschef</em> to run again</sub>

<!-- gh-aw-agentic-workflow: PR Sous Chef, engine: pi, model: copilot/gpt-5.4, id: 30772836840, workflow_id: pr-sous-chef, run: https://github.com/github/gh-aw/actions/runs/30772836840 -->

[pelikhan]

@copilot use inlined skills instead

[pelikhan]

@copilot update ambient context optimizer and token optimization instructions to recommend inlined skills to delay load prompt instructions

[Copilot]

@copilot use inlined skills instead

Done. Replaced the separate .github/aw/cache-strategy-templates.md file with two inline ## skill: blocks (cache-strategy-issue-template and cache-strategy-discussion-template) at the end of the workflow. The gh-aw runtime extracts inline skills before the main prompt runs, keeping them out of the initial request while still available on demand.

[Copilot]

@copilot update ambient context optimizer and token optimization instructions to recommend inlined skills to delay load prompt instructions

Done. Two files updated:

• .github/aw/token-optimization.md: expanded the Quick-Reference "Sub-skills" checklist item to explicitly call out inline ## skill: blocks and delayed loading; added Technique 3b with when-to-use guidance, a pattern example, and a note on when to use a shared import vs. an inline skill.
• .github/workflows/daily-ambient-context-optimizer.md: added detection of large inline output templates as a Source Review signal, and added priority 6 to the Recommendation Rules to suggest moving such templates to ## skill: blocks.

[Copilot]

Pull request overview

Reduces the cache analyzer’s initial prompt by moving output templates into on-demand inline skills.

Changes:

• Converts issue and discussion templates into inline skills.
• Updates optimizer guidance and token-optimization documentation.
• Recompiles affected workflow lock files.

Show a summary per file

File	Description
.github/workflows/daily-cache-strategy-analyzer.md	Adds on-demand template skills.
.github/workflows/daily-cache-strategy-analyzer.lock.yml	Recompiled analyzer workflow.
.github/workflows/daily-ambient-context-optimizer.md	Recommends inline template skills.
.github/workflows/daily-ambient-context-optimizer.lock.yml	Recompiled optimizer workflow.
.github/aw/token-optimization.md	Documents delayed instruction loading.

Review details

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 5/5 changed files
• Comments generated: 3
• Review effort level: Balanced

[gh-aw-bot]

@copilot Please take a quick pass on the remaining PR hygiene items, then run the pr-finisher skill.

Notes to check before merging:

• The branch has been queued for refresh from the base branch.
• Confirm the workflow/doc wording consistently recommends inline ## skill: blocks for delayed-load templates, especially where the PR reviewer summary still refers to a shared file.
• Revisit the earlier agent (PR Code Quality Reviewer) run if it remains relevant after refresh: https://github.com/github/gh-aw/actions/runs/30771883818/job/91560725634

If everything still looks good after refresh, please wrap up with the pr-finisher skill.

Generated by 👨‍🍳 PR Sous Chef · gpt54 · 16.7 AIC · ⌖ 6.74 AIC · ⊞ 8.3K · ◷
_{Comment /souschef to run again}

[github-actions]

🎉 This pull request is included in a new release.

Release: v0.84.3