Harden Daily Ambient Context Optimizer issue publication against placeholder and oversize payload failures

[Copilot]

Daily Ambient Context Optimizer could enter a dead-end failure path by calling create_issue with a placeholder body first, then attempting a payload above the safe-output 10KB limit. This change tightens prompt constraints and allows bounded retries so a single invalid call does not consume the workflow’s only issue-creation attempt.

• Failure-mode hardening

  • Increased safe-outputs.create-issue.max from 1 to 3 for this workflow so invalid first attempts do not block recovery.
  • Preserves the workflow contract of producing one real report while tolerating transient tool-call mistakes.

• Prompt guardrails for issue creation

  • Added explicit prohibition on placeholder/probe/test create_issue calls (e.g. ".", "test").
  • Added explicit sequencing rule: call create_issue only after validating final body content and length.

• Body-size control + retry behavior

  • Added a concrete size target (< 9500 UTF-8 bytes) before submission to stay below safe-output limits.
  • Added fallback guidance: if body-size validation fails, retry with a compact body that preserves core sections (Executive Summary, Highest-Leverage Changes, Key Metrics, References).

safe-outputs:
  create-issue:
    max: 3

# prompt guardrails
- Keep the final issue body under 9500 bytes (UTF-8)...
- Never call `create_issue` with placeholders...
- Call `create_issue` only after validating your final body length and content.

[github-actions]

Hey @Copilot 👋 — thanks for hardening the Daily Ambient Context Optimizer against placeholder calls and oversized payloads. The three-pronged approach (retry budget, prompt guardrails, and body-size target) is well-reasoned and the PR description clearly documents the failure mode being addressed.

One thing that would strengthen this further:

• Test coverage — there are no changes to test files alongside the prompt/safe-output config changes. Even for workflow prompt files, a lightweight integration test or a snapshot assertion on the rendered safe-outputs config (e.g. asserting max: 3 and the new prompt lines appear in the compiled workflow) would give the team confidence this doesn't regress. Check whether the repo's test suite has a pattern for locking or snapshotting workflow configs.

If you'd like a hand adding coverage, you can assign this prompt to your coding agent:

Add a test (or update the existing snapshot/lock test) for the daily-ambient-context-optimizer workflow to assert:
1. The safe-outputs config for `create_issue` has `max: 3`.
2. The compiled prompt includes the guardrail lines: 'Never call `create_issue` with placeholders' and 'Keep the final issue body under 9500 bytes'.
3. The fallback retry instruction ('shorten the details and retry') is present in the workflow body.
Use whatever test pattern already exists in the repo for compiled workflow validation.

Generated by ✅ Contribution Check · 153 AIC · ⌖ 7.31 AIC · ⊞ 5.9K · ◷

[Copilot]

Pull request overview

This PR hardens the Daily Ambient Context Optimizer workflow’s issue-publication path so that placeholder/invalid create_issue calls and oversize bodies don’t trap the run in a dead-end failure (while still producing exactly one real report).

Changes:

• Increased safe-outputs.create-issue.max from 1 → 3 to permit bounded retries after invalid issue-creation attempts.
• Tightened prompt guardrails to forbid placeholder/probe create_issue calls and to require validating final body content + size (targeting < 9500 UTF-8 bytes) before submission.
• Added explicit retry guidance for body-size validation failures (shorten and retry with a compact, section-preserving report).

Show a summary per file

File	Description
.github/workflows/daily-ambient-context-optimizer.md	Updates safe-output allowance and strengthens prompt sequencing/size/retry guardrails for create_issue.
.github/workflows/daily-ambient-context-optimizer.lock.yml	Recompiled workflow output reflecting the updated safe-output constraints and prompt content.

Copilot's findings

Tip

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

• Files reviewed: 2/2 changed files
• Comments generated: 1