fix(submit): report failure when registry rejects submission

[knoxgraeme]

Summary

Fixes #68

When submitSkillsToRegistry returns failed submissions (e.g. HTTP 429 non-JSON from the registry), the JSON output incorrectly reported success: true with exit_code: 0. Automation relying on these signals would silently miss failed submissions.

Root cause (src/commands/submit.ts): the else branch that pushes to failed[] never set jsonOutput.success = false, and the JSON exit always called outputJsonAndExit(EXIT_CODES.SUCCESS) unconditionally.

Changes:

• Set jsonOutput.success = false when any submission lands in failed[]
• Exit with EXIT_CODES.NETWORK_ERROR (3) in both JSON and TTY modes when the registry rejects the submission
• Add src/__tests__/submit.test.ts covering argument validation and the JSON failure contract

Before:

{ "success": true, "failed": [{ "skill": "...", "reason": "Registry returned non-JSON response (HTTP 429): ..." }], "exit_code": 0 }

After:

{ "success": false, "failed": [{ "skill": "...", "reason": "Registry returned non-JSON response (HTTP 429): ..." }], "exit_code": 3 }

Test plan

• All 7 new submit.test.ts tests pass
• success: false + non-zero exit_code for invalid args in JSON mode
• Non-zero exit for non-existent repo (covers failure path before registry)
• Existing test suite unaffected

https://claude.ai/code/session_01CogjFAAGo379Vv3dPLPkzP

---

Generated by Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 06039a0602

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".

[chatgpt-codex-connector]

Avoid live GitHub calls in submit tests

This test invokes the real CLI with a syntactically valid repo, so it proceeds into discovery and hits GitHub (fetchDefaultBranch calls https://api.github.com/repos/...). In CI runs without network, behind a proxy, or when GitHub rate-limits the runner, the test can fail or time out even though the JSON failure contract is unchanged; mock the discovery/registry path instead of depending on a live nonexistent repository.

Useful? React with 👍 / 👎.

[knoxgraeme]

The concern is valid, but the existing test infrastructure (invokeCli via execFileSync) spawns a separate process, so vi.mock can't intercept network calls made inside it. Adding proper mocking would require a different test architecture (e.g. an in-process test runner or an HTTP interceptor like MSW). The same live-call pattern already exists in add.test.ts (line 33–38). Happy to track adding proper test mocking infrastructure as a follow-up improvement rather than blocking this fix.

---

Generated by Claude Code

[caioribeiroclw-pixel]

Thanks — this covers the failure semantics I hit in #68 from the reporter side.

I checked this branch locally and the important behavior is now aligned with what agents/CI need: JSON mode does not report a failed submission as success: true, and the command exits non-zero when the submission path fails.

Validation I ran on 06039a0:

• npm run typecheck
• npm test → 242/242 passing
• npm run build
• git diff --check

I’m going to close my duplicate PR #69 so review can converge here.