[teamleaderleo]

Leo-ask: ask-973a915bea

Leo, 2026-10-28: "Where the f***ing are the generalized solution to all these problems that these incessant pins?"

This is step 1 of 4 on feat-cmux-next. It is groundwork and changes no workflow. 475 runs-on lines end in a literal single-size Blacksmith label, and a re-run cannot move a literal. The next PRs point each of those at its size's pool, fromJSON(vars.RUNNER_POOLS).<pool>. main declares those pools (#32393), and runner_pools.py keeps them on a live member. The last PR adds a guard test that refuses a new literal.

What this PR changes:

• .github/runners.json gets a copy of main's pools declaration. tests/test_ci_runner_pools_copy.py keeps the copy equal to main's wherever origin/main is in the checkout. It also refuses a workflow that reads an undeclared pool, which would read '' and queue forever.
• tests/runs_on_labels.py now parses fromJSON(vars.RUNNER_POOLS).<pool> and resolves it to that pool's members (never ''). Every guard built on it keeps resolving labels after a literal becomes a pool read. Before this change, the tokenizer stopped at .<pool>.
• docs/ci-runners.md gets a Runner pools section.

Observed / Not run

• Ran python3 tests/X.py for every test that imports runs_on_labels, plus test_ci_runner_pools_copy, test_ci_runner_set, test_ci_repo_variable_defaults, test_ci_fork_runner_routing and test_ios_internal_builder. All pass.
• Not run: the whole tooling-tests lane.

Performance

metric	before	after	delta	harness
Blacksmith minutes per day	unchanged	unchanged	0	tests and docs only; no runs-on changes here

Changelog

none

Agent session: agent session

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Makes CI tests resolve fromJSON(vars.RUNNER_POOLS).<pool> reads to the declared pool members, so label guards keep working when a literal runs-on label becomes a pool read. Workflows themselves are unchanged.

• Copies main's pools declaration into .github/runners.json; a new hand-run test keeps the copy equal to main's across the full pool set and refuses reads of an undeclared pool (which would resolve to '' and queue forever).
• Extends tests/runs_on_labels.py's parser with property access (.pool) and resolves it to that pool's members.
• Adds a Runner pools section to docs/ci-runners.md.

^{Written for commit d570ad0. Summary will update on new commits.}

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review d570ad0: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr32417 -f ref=lane/runner-pools-resolve -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4147460604401 (87f859ba07)
• cmux NEXT nightly 1.0.0-nightly.4147660705901 (b27c0b3daf)
• cmux NEXT nightly 1.0.0-nightly.4148000752901 (17b75935f0)

[blacksmith-sh]

Found 4 test failures on Blacksmith runners:

Failures

Test	View Logs
main.DeveloperIdAppTests.test_download_prints_the_active_profile_only/ test_download_prints_the_active_profile_only	View Logs
stubs: /System/Library/Frameworks/Security.framework/Versions/A/Security has no stub/ Security has no stub	View Logs
stubs: CoreFoundation.tbd: install-name /System/Library/Frameworks/ CoreFoundation.framework/Versions/A/CoreFoundation, missing 1: ['_CFRetain']/ CoreFoundation, missing 1: ['_CFRetain']	View Logs
stubs: CoreFoundation.tbd: install-name /System/Library/Frameworks/ CoreFoundation.framework/Versions/A/CoreFoundation, missing 1: ['_log2']/ CoreFoundation, missing 1: ['_log2']	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[teamleaderleo]

Review (subagent claude-opus-5-5 medium): SHIP at d570ad0

The earlier blocker is fixed: the pools copy test is a manual test with a reason and is listed in UNWIRED, and the wiring test passes. The tokenizer was checked over all 2,854 workflow expressions: none tokenize differently, and the 17 that newly parse are all fromJSON(...).x reads. The pools copy now equals main's 15 pools. The registry test's git-fixture error is environmental and fails the same way on base.

Follow-ups: the fromJSON(vars.RUNNER_POOLS || '{}').x form stays unresolved; no unit test for the prop node yet; $comment indentation in runners.json.

[teamleaderleo]

Review (codex gpt-6.1-sol high, sweep, via the review router; full diff from merge base 9508b70 to this head): SHIP at cf950a4

No P1/P2 findings.