Skip to content

fix(skills): macOS verification routing and foundation entry point - #1

Merged
brbndon merged 5 commits into
mainfrom
cursor/macos-verification-routing-dea4
Jul 23, 2026
Merged

fix(skills): macOS verification routing and foundation entry point#1
brbndon merged 5 commits into
mainfrom
cursor/macos-verification-routing-dea4

Conversation

@brbndon

@brbndon brbndon commented Jul 21, 2026

Copy link
Copy Markdown
Owner

Summary

  • Fixes iOS-simulator bias in bootstrap/verification: macOS apps use the XcodeBuildMCP macos workflow (build_run_macos / test_macos); ui-automation remains iOS-only.
  • Strengthens $apple-development-foundation as the global entry point for new-app bootstrap routing (prefer over generic macOS/iOS skills).
  • Adds a competing-skills routing plan and evaluation fixtures for macOS bootstrap.

Verification

  • ./Scripts/generate-skill-catalog.py --check
  • ./Scripts/validate-skill-evaluations.py
  • ./Scripts/verify-skills.sh
  • ./Scripts/test-install-skills.sh

Notes

XcodeBuildMCP docs confirm test_macos exists; ui-automation is iOS-simulator-only, so macOS verification requires tests (or an explicit skip + residual risk), not screenshots.

… entry

- Branch bootstrap/verification to macos tools (test_macos) vs iOS simulator
- Strengthen apple-development-foundation as global bootstrap entry point
- Add competing macOS skills routing plan and evaluation fixtures

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d6bb593143

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread README.md
Point Codex at this repo when bootstrapping apps:

> Use the skills from `/path/to/AppleDevelopmentFoundation/.agents/skills` (start with `codex-bootstrap`).
> Use the skills from `/path/to/AppleDevelopmentFoundation/.agents/skills` (start with `$apple-development-foundation` → `$codex-bootstrap` for new apps).

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Route every new-app entry point through the foundation

This recommendation is not reflected in the other primary entry pages: MCP.md:65-67, docs/index.mdx:27, and docs/skills/index.mdx:44 still tell a fresh user to start directly with $codex-bootstrap. Users entering through those documented paths therefore bypass the router despite the new global-entry contract; update those entry points, or qualify direct bootstrap as only for an already-selected skill, so routing is deterministic.

AGENTS.md reference: AGENTS.md:L15-L15

Useful? React with 👍 / 👎.

brbndon added 2 commits July 21, 2026 12:26
- Point MCP and primary docs at $apple-development-foundation first
- Qualify direct $codex-bootstrap as only when already selected
- Require foundation → bootstrap for new-app recipes and consumer bootstrap
- Qualify direct child invocation in quickstart

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7026f772ad

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment on lines +100 to +104
- **macOS:** use the `macos` workflow — prefer `build_run_macos` for a launch
smoke when useful; run `test_macos` when a test target exists (required
verification when tests are present). Default
`extraArgs: ["-parallel-testing-enabled", "NO"]`. Do **not** call
`test_sim`, `build_run_sim`, or ui-automation screenshot/hierarchy tools —

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Update generated consumer guidance for macOS verification

When a macOS-only bootstrap creates the recommended consumer guidance, assets/consumer-AGENTS.md.template:95-115 still mandates simulator management, an exact simulatorId, build_run_sim, and test_sim. That scoped AGENTS.md then governs future work in the consumer repository and conflicts with this new macOS branch, directing agents to iOS-only tools instead of build_run_macos and test_macos; branch the generated guidance by platform as part of this change.

Useful? React with 👍 / 👎.

Comment on lines +133 to +135
"skill_ids": [
"codex-bootstrap",
"swift-testing-verification"

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Include the foundation router in the bootstrap fixture

This prompt does not indicate that $codex-bootstrap has already been selected, but its expected ordered shortlist omits the newly required $apple-development-foundation entry point. A model following the repository's new-app routing contract would therefore fail this prompt-conformance fixture; prepend apple-development-foundation, or explicitly state in the prompt that the child skill was already selected.

AGENTS.md reference: AGENTS.md:L15-L15

Useful? React with 👍 / 👎.

- Branch generated consumer XcodeBuildMCP guidance for iOS vs macOS
- Prepend apple-development-foundation on new-app bootstrap fixtures
- Assert macOS verification tools in consumer guidance tests

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: bef377a903

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread .agents/skills/codex-bootstrap/SKILL.md Outdated
Comment on lines +100 to +102
- **macOS:** use the `macos` workflow — prefer `build_run_macos` for a launch
smoke when useful; run `test_macos` when a test target exists (required
verification when tests are present). Default

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Require at least one macOS build when tests are absent

For a macOS-only bootstrap with no test target, this branch makes build_run_macos optional (“when useful”) and explicitly permits skipping test_macos, so the workflow can declare the generated skeleton complete without ever compiling it. This also conflicts with the bootstrap checklist in MCP.md:82, which includes build_run_macos; require at least build_macos when a launch smoke or tests are not run.

AGENTS.md reference: AGENTS.md:L13-L13

Useful? React with 👍 / 👎.

- Prefer build_run_macos / test_macos, but require at least build_macos
  when neither launch smoke nor tests run
- Align MCP checklist, consumer AGENTS template, and docs; assert build_macos
  in consumer guidance tests

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 84d9cc8a49

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

2. discover_projs only if defaults still wrong/missing
3. build_run_sim for launch (or build_run_macos / build_run_device)
4. test_sim / test_macos if present
3. iOS: build_run_sim → test_sim; macOS: enable macos workflow → build_run_macos (or at least build_macos if no launch/tests) → test_macos if present (no ui-automation)

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Keep the physical-device branch in the verification prompt

When this generic prompt is used to verify an iOS feature on a physical device, the new platform branch now sends every iOS target through build_run_sim / test_sim and drops the previous build_run_device option. Physical-device support still exists and is documented as build_run_device / test_device in MCP.md:46 and docs/tools/xcodebuildmcp.mdx:138-139, so add a device-workflow branch (and accept a device target in Inputs) rather than silently verifying the wrong device class.

Useful? React with 👍 / 👎.

@brbndon
brbndon merged commit c5eeebe into main Jul 23, 2026
2 checks passed
@brbndon
brbndon deleted the cursor/macos-verification-routing-dea4 branch July 23, 2026 17:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant