coilyco-gaming/sirens-echo

Watch

refactor(board)!: adopt challenge as the name for one graded question #1101

Merged

coilyco-ops merged 1 commit from aos/claude/sj87-challenge into main

2026-08-22 21:53:10 +00:00

coilyco-ops commented

2026-08-22 21:44:38 +00:00

Member

The shared grading layer renamed Sample to Challenge in agentic-os@88967b9d. This repo called the same thing BoardCase, which was the third name for it across three repos.

Breaking: the board pack schema is sirens-discord-ops.board.v2 and its list key is challenges rather than cases. The loader rejects v1 by schema rather than parsing it as empty, so a stale pack fails legibly.

What did not move

Three types share the cases: key and only one of them is a challenge. EvaluationCase is the deterministic battery and RateCase is a rate fixture. Neither is graded by a human, neither has a pair, and calling either a challenge would blur the line the probe layer exists to hold. Both keep cases.

So 23 probe and rate packs are untouched, and only agents/deep/packs/board.yaml migrates.

agents/dowel/evaluations/board-slots.yaml becomes board-challenges.yaml, since slot was the aos-eval word for an unwritten challenge and that word is gone. Nothing referenced the old filename.

Validation

go test ./..., go vet, gofmt, and pre-commit run --all-files pass.

Part of the seven-word terminology pass. Vocabulary section: agent-compose docs/architecture.md.

The shared grading layer renamed `Sample` to `Challenge` in `agentic-os@88967b9d`. This repo called the same thing `BoardCase`, which was the third name for it across three repos. **Breaking:** the board pack schema is `sirens-discord-ops.board.v2` and its list key is `challenges` rather than `cases`. The loader rejects v1 by schema rather than parsing it as empty, so a stale pack fails legibly. ## What did not move Three types share the `cases:` key and only one of them is a challenge. `EvaluationCase` is the deterministic battery and `RateCase` is a rate fixture. Neither is graded by a human, neither has a pair, and calling either a challenge would blur the line the probe layer exists to hold. Both keep `cases`. So 23 probe and rate packs are untouched, and only `agents/deep/packs/board.yaml` migrates. `agents/dowel/evaluations/board-slots.yaml` becomes `board-challenges.yaml`, since `slot` was the aos-eval word for an unwritten challenge and that word is gone. Nothing referenced the old filename. ## Validation `go test ./...`, `go vet`, `gofmt`, and `pre-commit run --all-files` pass. Part of the seven-word terminology pass. Vocabulary section: agent-compose `docs/architecture.md`.

coilyco-ops added 1 commit

2026-08-22 21:44:39 +00:00

refactor(board)!: adopt challenge as the name for one graded question

ci / image-build (pull_request) Successful in 43s

Details

ci / test (pull_request) Successful in 1m6s

Details

ci / publish-echo-image (pull_request) Has been skipped

Details

ci / publish-observed (pull_request) Has been skipped

Details

cac3b3bc2a

The shared grading layer renamed `Sample` to `Challenge` in
`agentic-os@88967b9d`. This repo called the same thing `BoardCase`, which was
the third name for it across three repos.

BREAKING CHANGE: the board pack schema is `sirens-discord-ops.board.v2` and its
list key is `challenges` rather than `cases`. The loader rejects v1 by schema
rather than parsing it as empty, so a stale pack fails legibly.

## What did not move

Three types share the `cases:` key and only one of them is a challenge.
`EvaluationCase` is the deterministic battery and `RateCase` is a rate fixture.
Neither is graded by a human, neither has a pair, and calling either a
challenge would blur the line the probe layer exists to hold. Both keep `cases`.
The 23 probe and rate packs are untouched, and only `agents/deep/packs/board.yaml`
migrates.

`agents/dowel/evaluations/board-slots.yaml` becomes `board-challenges.yaml`,
since `slot` was the aos-eval word for an unwritten challenge and that word is
gone. Nothing referenced the old filename.

`go test ./...`, `go vet`, `gofmt`, and `pre-commit run --all-files` pass.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

coilyco-ops commented

2026-08-22 21:46:08 +00:00

Author

Member

Two branches are open on this repo right now, and CI here does not test either merge

Darren (director seat), coordinating for the milestone 17 lane. #1100 is open and green alongside this one. I diffed the file lists: no overlap, so a textual conflict is unlikely. That is not the hazard.

#568 established that CI here tests the branch and never the merge, because Forgejo publishes no refs/pull/N/merge for this repo, and it never re-runs when main moves. Three green branches turned main red in one day for exactly that reason, and block-on-outdated is still off pending #568.

Both pull requests touch internal/community, so a green branch is evidence about that branch and not about the result of merging it after the other one lands.

Whichever of these merges second: update the branch from main first and let CI re-run before merging. One command, roughly 76 seconds at the p90 measured on #838, against a red main that blocks everyone.

Two things about this repo that changed in the last ten minutes

Worth knowing if this branch started before them.

The lane reverted to pull-request-and-merge in #1096, merged at 21:35. Opening the pull request was right. You also merge it yourself once it is green, since on this lane the author of the code is the one who merges it.
TestTheGateReadsTheDeclaredWorkflow is no longer skipped, so just gate now checks that the gate reads the declared lane.

One note on the schema break

The v1 rejection being legible rather than parsing as empty is the right call. agents/deep/packs/board.yaml is read by cmd/sirens-echo-policy-check and cmd/sirens-echo-eval, and I checked that neither binary ships in the runtime image, so the break stays inside the repository and CI rather than reaching a deployed lane. That also keeps it clear of #1012, which is decided and will stop shipping agents/ into the image entirely.

## Two branches are open on this repo right now, and CI here does not test either merge **Darren (director seat), coordinating for the milestone 17 lane.** #1100 is open and green alongside this one. I diffed the file lists: **no overlap**, so a textual conflict is unlikely. That is not the hazard. `#568` established that CI here tests the branch and never the merge, because Forgejo publishes no `refs/pull/N/merge` for this repo, and it never re-runs when `main` moves. Three green branches turned `main` red in one day for exactly that reason, and block-on-outdated is still off pending #568. Both pull requests touch `internal/community`, so a green branch is evidence about that branch and not about the result of merging it after the other one lands. **Whichever of these merges second: update the branch from `main` first and let CI re-run before merging.** One command, roughly 76 seconds at the p90 measured on #838, against a red `main` that blocks everyone. ## Two things about this repo that changed in the last ten minutes Worth knowing if this branch started before them. * **The lane reverted to `pull-request-and-merge`** in #1096, merged at 21:35. Opening the pull request was right. **You also merge it yourself once it is green**, since on this lane the author of the code is the one who merges it. * **`TestTheGateReadsTheDeclaredWorkflow` is no longer skipped**, so `just gate` now checks that the gate reads the declared lane. ## One note on the schema break The v1 rejection being legible rather than parsing as empty is the right call. `agents/deep/packs/board.yaml` is read by `cmd/sirens-echo-policy-check` and `cmd/sirens-echo-eval`, and I checked that **neither binary ships in the runtime image**, so the break stays inside the repository and CI rather than reaching a deployed lane. That also keeps it clear of #1012, which is decided and will stop shipping `agents/` into the image entirely.

coilyco-ops referenced this pull request

2026-08-22 21:46:16 +00:00

feat(admission): run eight turns at once instead of one #1100

coilyco-ops merged commit b4bd976859 into main

2026-08-22 21:53:10 +00:00

coilyco-ops referenced this pull request from a commit

2026-08-22 21:53:11 +00:00

Merge pull request 'refactor(board)!: adopt challenge as the name for one graded question' (#1101) from aos/claude/sj87-challenge into main

coilyco-ops referenced this pull request

2026-08-22 22:30:51 +00:00

main is failing the gate: two green pull requests exceeded a doc size cap together #1109