Codex collector: plan label and model-scoped limit windows

[nixfred]

Changes

Brings the Codex tab of the Agents panel up to parity with the Claude tab. account/rateLimits/read already returns more than the collector showed, so this only surfaces what Codex reports:

• tierLabel becomes a human plan name (prolite → "Pro Lite", plus → "Plus") instead of the raw API token, matching Claude's "Max 20x" / "Pro".
• Window labels follow the Claude collector's wording ("Session (5-hour)", "Weekly (7-day)") so the LIMITS section reads the same on both tabs.
• Model-scoped allowances under rateLimitsByLimitId (e.g. "GPT-5.3-Codex-Spark", each with its own 5-hour and weekly window) are emitted with a title, the way Claude reports "Fable Weekly", so the panel renders them as their own rows instead of dropping them.
• spendControlReached surfaces as the usage status text.
• Launches app-server with -a on-request, same as Fix Codex usage collection on 0.149 #7649 (Codex 0.149 retired untrusted, Codex 0.149 breaks Agents panel limit display #7648). Happy to rebase on Fix Codex usage collection on 0.149 #7649 once it lands — this is meant to stack on it, not compete with it.

Before (Codex 0.149.0, Pro Lite): no plan, limits: [].
After:

tierLabel: "Pro Lite"
Weekly (7-day)                          6%
GPT-5.3-Codex-Spark Session (5-hour)    0%
GPT-5.3-Codex-Spark Weekly (7-day)      0%

Tests

• bash test/shell.d/agent-usage-codex-scanner-test.sh — mock app-server now answers with a realistic rateLimits / rateLimitsByLimitId payload; asserts the plan label and every limit row.
• ./test/cli
• live bin/omarchy-agent-usage-codex --limits-only against Codex CLI 0.149.0 on Omarchy 4.0.0-1.

🤖 Generated with Claude Code

https://claude.ai/code/session_01PDPn6hVmawjV1sSRLedsUD

[Copilot]

Pull request overview

Brings Codex usage reporting closer to Claude’s Agents panel presentation.

Changes:

• Adds human-readable plan and limit labels.
• Exposes model-scoped limits and spend-control status.
• Uses Codex’s supported on-request approval policy and expands tests.

Tip

If you aren't ready for review, convert to a draft PR.
Click "Convert to draft" or run gh pr ready --undo.
Click "Ready for review" or run gh pr ready to reengage.

Reviewed changes

Copilot reviewed 1 out of 2 changed files in this pull request and generated 1 comment.

File	Description
bin/omarchy-agent-usage-codex	Normalizes and collects expanded Codex limit data.
test/shell.d/agent-usage-codex-scanner-test.sh	Adds realistic plan and scoped-limit assertions.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[nixfred]

Rebased onto current quattro — sorry for the noise, and thank you for your patience on this one.

This PR had drifted so far from its base that GitHub was rendering it as ~118,000 lines across 100 files, which made the Codex collector change impossible to actually review. That was an artifact of the branch point, not the change.

It is now the real diff and nothing else. Tests pass on today's quattro.

No behaviour has changed from what was originally proposed — only the base it sits on.

[nixfred]

On the Copilot note about the mock not inspecting its arguments — that was right, and it is already handled in this branch. test/shell.d/agent-usage-codex-scanner-test.sh now rejects any invocation other than the expected one, so a regression back to -a untrusted fails the test rather than passing it:

# Fail if the collector regresses to `-a untrusted` (Codex 0.149 / #7648).
if [[ $1 != "-s" || $2 != "read-only" || $3 != "-a" || $4 != "on-request" || ${5:-} != "app-server" ]]; then
  echo "unexpected codex invocation: $*" >&2
  exit 1
fi

The comment predates the rebase, so it may be pointing at the older diff.