Smoke Copilot: enforce add_comment budget in prompt before output phase

[Copilot]

Smoke Copilot was intermittently failing late because the prompt could ask for more add_comment calls than safe-outputs.add-comment.max: 2 allows. This change moves the constraint into explicit early planning instructions so the model budgets comment calls up front instead of failing at safe-output processing time.

• Prompt-level guardrail (primary fix)

  • Added a new Hard Limit: add_comment Budget section to smoke-copilot.md.
  • Explicitly defines the two-call budget and allocates calls by event type to prevent overproduction of comments.

• Output contract tightened

  • Made output comment steps mutually exclusive:
    • pull_request runs: post the brief PR summary comment.
    • non-pull_request runs: post the fun discussion comment.
  • This removes the prior ambiguity where both could be produced in the same run.

• Golden artifacts updated

  • Synced the same constraint language and conditional output rules in:
    • pkg/workflow/testdata/wasm_golden/fixtures/smoke-copilot.md
    • pkg/workflow/testdata/wasm_golden/WasmBinary/smoke-copilot.golden
    • pkg/workflow/testdata/TestWasmGolden_CompileFixtures/smoke-copilot.golden

## Hard Limit: `add_comment` Budget

`safe-outputs.add-comment.max` is `2`. Never exceed 2 total `add_comment` calls in this run.

- Call #1 is required in discussion interaction testing.
- Call #2 depends on trigger:
  - `pull_request` event: post PR summary comment; skip discussion follow-up comment.
  - non-`pull_request` event: skip PR summary comment; post discussion follow-up comment.

[mnkiefer]

@copilot Recompile worklfows

[Copilot]

@copilot Recompile worklfows

I recompiled smoke-copilot (gh aw compile smoke-copilot --validate --verbose) and there were no resulting diffs, so the workflow lockfile is already up to date at a0f8760.

[github-actions]

⚠️ PR Code Quality Reviewer failed during code quality review.

[github-actions]

🧠 Matt Pocock Skills Reviewer has completed the skills-based review. ✅

[github-actions]

✅ Design Decision Gate 🏗️ completed the design decision gate check.

No ADR enforcement needed: PR #34611 does not have the 'implementation' label and has only 33 new lines of code in business logic directories (threshold: 100).

[github-actions]

🧪 Test Quality Sentinel completed test quality analysis.

No test files were added or modified in this PR. The only changes are to workflow markdown (.md) files and golden test fixtures (.golden). Test Quality Sentinel skipped.

[Copilot]

Pull request overview

This PR updates the Smoke Copilot prompt/test artifacts to explicitly budget add_comment safe-output calls up front (matching safe-outputs.add-comment.max: 2) and to make the two “extra comment” outputs mutually exclusive based on event type, preventing late safe-output failures caused by overproducing add_comment calls.

Changes:

• Added a new “Hard Limit: add_comment Budget” section that explicitly caps total add_comment calls at 2 and pre-allocates them.
• Tightened the Output contract so only one of the PR-summary comment vs. discussion follow-up comment can be produced (depending on pull_request vs. non-pull_request triggers).
• Updated wasm golden artifacts and fixture markdown to match the new prompt rules.

Show a summary per file

File	Description
pkg/workflow/testdata/wasm_golden/WasmBinary/smoke-copilot.golden	Adds explicit add_comment budgeting and makes Output comment steps conditional by event type.
pkg/workflow/testdata/wasm_golden/fixtures/smoke-copilot.md	Mirrors the new hard limit section and conditional Output instructions in the fixture prompt.
pkg/workflow/testdata/TestWasmGolden_CompileFixtures/smoke-copilot.golden	Updates compiled golden output to reflect the new budgeting/conditional Output contract.
.github/workflows/smoke-copilot.md	Adds the add_comment budget guardrail and makes the discussion Output comment conditional for non-PR runs.

Copilot's findings

Tip

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

• Files reviewed: 4/4 changed files
• Comments generated: 0

[github-actions]

🧠 Reviewed using Matt Pocock's skills by Matt Pocock Skills Reviewer · sonnet46 1.2M

[mnkiefer]

@copilot Implement suggested fix by github-actions.

[Copilot]

@copilot Implement suggested fix by github-actions.

Implemented in 4e0eb1f. I applied the github-actions suggestion by replacing brittle step-number references with semantic wording in the add_comment budget section, and synced the matching golden artifacts.