coilyco-flight-deck/ward

Watch

Describe the shipped exact-commit QA gate in engineer outcomes #1577

New issue

Closed

opened 2026-07-28 02:46:58 +00:00 by coilyco-ops · 5 comments

coilyco-ops commented

2026-07-28 02:46:58 +00:00

Member

Outcome

Ward's engineer seed and final outcome instructions must describe the shipped QA boundary accurately. They must not say brokered QA is pending after exact-commit QA has shipped.

Parent goal: #1574

Current evidence

Current cmd/ward/agent.go emits this instruction into every engineer run:

The in-container review gate was intentionally skipped because the temporary ward default pending brokered QA

The same framing is preserved in tests. Ward issue #844 already shipped the separate QA role and director merge checks that bind a structured QA verdict to the exact current pull request commit.

A live issue-scoped dry run for the parent goal reproduced the obsolete sentence on 2026-07-27.

Required behavior

Ward must describe QA as a separate, opt-in, exact-commit verification role.
Ward must explain that the in-container review gate is skipped because implementation and QA are role-separated, not because brokered QA is unfinished.
Pull-request-and-merge guidance must continue to require the current commit's green CI and trusted QA evidence under the actor-admission work.
Engineering must update the source tests or golden prompts that pin the old sentence.
Historical issue comments must remain unchanged.

Acceptance criteria

A generated engineer prompt contains no active pending brokered QA claim.
A generated engineer outcome instruction says that QA is separate, opt-in, and bound to the exact candidate commit.
Tests cover the wording in pull-request and pull-request-and-merge modes where it affects the final outcome contract.
Current QA and director merge behavior remains unchanged.
Ward's repository-declared vet, lint, test, and pre-commit validation passes.

docs/FEATURES.md does not need a wording-only update.

Codex, via advisor surface

## Outcome Ward's engineer seed and final outcome instructions must describe the shipped QA boundary accurately. They must not say brokered QA is pending after exact-commit QA has shipped. Parent goal: https://forgejo.coilysiren.me/coilyco-flight-deck/ward/issues/1574 ## Current evidence Current `cmd/ward/agent.go` emits this instruction into every engineer run: ```text The in-container review gate was intentionally skipped because the temporary ward default pending brokered QA ``` The same framing is preserved in tests. Ward issue https://forgejo.coilysiren.me/coilyco-flight-deck/ward/issues/844 already shipped the separate QA role and director merge checks that bind a structured QA verdict to the exact current pull request commit. A live issue-scoped dry run for the parent goal reproduced the obsolete sentence on 2026-07-27. ## Required behavior * Ward must describe QA as a separate, opt-in, exact-commit verification role. * Ward must explain that the in-container review gate is skipped because implementation and QA are role-separated, not because brokered QA is unfinished. * Pull-request-and-merge guidance must continue to require the current commit's green CI and trusted QA evidence under the actor-admission work. * Engineering must update the source tests or golden prompts that pin the old sentence. * Historical issue comments must remain unchanged. ## Acceptance criteria * A generated engineer prompt contains no active `pending brokered QA` claim. * A generated engineer outcome instruction says that QA is separate, opt-in, and bound to the exact candidate commit. * Tests cover the wording in pull-request and pull-request-and-merge modes where it affects the final outcome contract. * Current QA and director merge behavior remains unchanged. * Ward's repository-declared vet, lint, test, and pre-commit validation passes. `docs/FEATURES.md` does not need a wording-only update.  Codex, via advisor surface

coilyco-ops referenced this issue

2026-07-28 02:47:45 +00:00

Stabilize Ward for safe autonomous burndown and verified releases #1574

coilyco-ops referenced this issue from a pull request that will close it,

2026-07-28 02:56:57 +00:00

Update engineer QA skip wording #1578

coilyco-ops commented

2026-07-28 02:57:16 +00:00

Author

Member

WARD-WORKFLOW: blocked 🛑

details

workflow: pull-request; review summary: blocked: repo-wide ward exec test fails in cmd/ward/container_hostnet_test.go (TestBuildUpPlanTailnet expects host-net on linux but got HostNet=false TSSidecar=true).
The wording fix is in place and the focused checks passed (go test ./cmd/ward -run 'TestReviewGateClauseInSeed|TestEngineerSeedDefaultsSkipReviewGate|TestAgentSeedPromptWorkflow|TestReviewGateWantedHonorsSkipsAndConfig', ward exec vet, ward exec lint, pre-commit run --all-files), but the branch cannot go green until that unrelated test failure is cleared.
The reservation remains held while this blocker is outstanding.

— Codex, via ward agent

WARD-WORKFLOW: blocked 🛑 <details><summary>details</summary> workflow: pull-request; review summary: blocked: repo-wide `ward exec test` fails in `cmd/ward/container_hostnet_test.go` (`TestBuildUpPlanTailnet` expects host-net on linux but got `HostNet=false TSSidecar=true`). The wording fix is in place and the focused checks passed (`go test ./cmd/ward -run 'TestReviewGateClauseInSeed|TestEngineerSeedDefaultsSkipReviewGate|TestAgentSeedPromptWorkflow|TestReviewGateWantedHonorsSkipsAndConfig'`, `ward exec vet`, `ward exec lint`, `pre-commit run --all-files`), but the branch cannot go green until that unrelated test failure is cleared. The reservation remains held while this blocker is outstanding. </details>  — Codex, via `ward agent`

coilyco-ops commented

2026-07-28 02:57:39 +00:00

Author

Member

WARD-WORKFLOW: reservation-released

release details

Run finished with WARD-WORKFLOW: blocked 🛑.

ward container reap released container engineer-codex-ward-1577 (--harness codex): the terminal outcome supersedes the reservation, so a later redispatch no longer needs --override-reservation.

— Codex, via ward agent

WARD-WORKFLOW: reservation-released <details><summary>release details</summary> Run finished with `WARD-WORKFLOW: blocked 🛑`. `ward container reap` released container `engineer-codex-ward-1577` (`--harness codex`): the terminal outcome supersedes the reservation, so a later redispatch no longer needs `--override-reservation`. </details>  — Codex, via `ward agent`

coilyco-ops commented

2026-07-28 03:04:41 +00:00

Author

Member

WARD-WORKFLOW: qa-blocked 🛑

qa details

verdict: blocked
reviewed_sha:
reviewer_family: internal
workflow: pull-request-and-merge
issue_ref: coilyco-flight-deck/ward#1577
pr_ref:
reason: ## Verdict
Blocked advisory: PR #1578 at 06b75094f92e4d54bbf1f08011f7da7929bc0c1f appears to implement the requested wording change correctly, and the live PR check is green, but the issue/PR thread on 2026-07-28 still records the branch as blocked by a repo-wide ward exec test failure outside this wording change.

What passed

The production prompt text in cmd/ward/agent.go now describes QA as role-separated, opt-in, and exact-commit-bound, and it no longer emits the stale pending brokered QA wording. The test updates cover both the default engineer skip path and the pull-request-and-merge skipped-review outcome path.

Why this is not a pass

Issue acceptance explicitly requires repo-declared vet, lint, test, and pre-commit validation to pass. The only live commit status exposed for this SHA is test / test (pull_request): success, but the authoritative issue/PR comments say full ward exec test was still failing in cmd/ward/container_hostnet_test.go. That leaves the acceptance state unresolved for this issue even though the wording fix itself looks correct.
evidence:

PR #1578 is open from branch issue-1577 into main, with head SHA 06b75094f92e4d54bbf1f08011f7da7929bc0c1f.
The diff is narrowly scoped to cmd/ward/agent.go, cmd/ward/agent_review_test.go, and cmd/ward/agent_workflow_test.go.
In cmd/ward/agent.go, temporaryReviewGateSkipReason now says QA is a separate, opt-in exact-commit verification role bound to the exact candidate commit.
In cmd/ward/agent.go, the skipped-review outcome text now says the in-container review gate was skipped because implementation and QA are role-separated, not because brokered QA is pending.
A tree grep on the candidate SHA found pending brokered QA only in negative test assertions, not in production prompt text.
The updated tests explicitly cover the engineer default skip wording and the pull-request-and-merge skipped-review path.
The issue comment posted on 2026-07-28T02:57:16Z reports: focused wording checks passed, but repo-wide ward exec test failed in cmd/ward/container_hostnet_test.go (TestBuildUpPlanTailnet).
The PR comment posted on 2026-07-28T02:57:16Z repeats the same blocker and says the branch cannot go green until that unrelated failure is cleared.
The live commit-status API for SHA 06b75094... reports one surfaced check: test / test (pull_request) = success at 2026-07-28T02:57:00Z.

risks:

The surfaced PR status and the issue-thread narrative disagree: the visible PR check is green, but the recorded repo-wide ward exec test result is still failing. That mismatch makes acceptance ambiguous.
Because this QA pass is read-only and the full repo test run is not exposed as a passing check for this SHA, I cannot confirm the final acceptance criterion from live state alone.

next steps:

Re-run or surface the full repo-declared validation for SHA 06b75094f92e4d54bbf1f08011f7da7929bc0c1f, especially ward exec test.
If TestBuildUpPlanTailnet is still failing, treat this issue as waiting on that external blocker rather than on the wording change itself.
If the full suite is now green, update the issue/PR thread so the authoritative record matches the live check state.

run_identity: KAI-DESKTOP-TOWER

dispatcher framing:
Read the issue title, body, and comment thread below as the QA brief. Inspect the candidate branch, any linked pull request, and the available checks in the live repository state. Return a structured QA verdict that a human can read at a glance. Do not edit files, commit, push, or otherwise change implementation state.

— Codex, via ward agent

WARD-WORKFLOW: qa-blocked 🛑 <details><summary>qa details</summary> verdict: blocked reviewed_sha: reviewer_family: internal workflow: pull-request-and-merge issue_ref: coilyco-flight-deck/ward#1577 pr_ref: reason: ## Verdict Blocked advisory: PR [#1578](https://forgejo.coilysiren.me/coilyco-flight-deck/ward/pulls/1578) at `06b75094f92e4d54bbf1f08011f7da7929bc0c1f` appears to implement the requested wording change correctly, and the live PR check is green, but the issue/PR thread on **2026-07-28** still records the branch as blocked by a repo-wide `ward exec test` failure outside this wording change. ## What passed The production prompt text in `cmd/ward/agent.go` now describes QA as role-separated, opt-in, and exact-commit-bound, and it no longer emits the stale `pending brokered QA` wording. The test updates cover both the default engineer skip path and the `pull-request-and-merge` skipped-review outcome path. ## Why this is not a pass Issue acceptance explicitly requires repo-declared `vet`, `lint`, `test`, and `pre-commit` validation to pass. The only live commit status exposed for this SHA is `test / test (pull_request): success`, but the authoritative issue/PR comments say full `ward exec test` was still failing in `cmd/ward/container_hostnet_test.go`. That leaves the acceptance state unresolved for this issue even though the wording fix itself looks correct. evidence: - PR `#1578` is open from branch `issue-1577` into `main`, with head SHA `06b75094f92e4d54bbf1f08011f7da7929bc0c1f`. - The diff is narrowly scoped to `cmd/ward/agent.go`, `cmd/ward/agent_review_test.go`, and `cmd/ward/agent_workflow_test.go`. - In `cmd/ward/agent.go`, `temporaryReviewGateSkipReason` now says QA is `a separate, opt-in exact-commit verification role bound to the exact candidate commit`. - In `cmd/ward/agent.go`, the skipped-review outcome text now says the in-container review gate was skipped because implementation and QA are role-separated, not because brokered QA is pending. - A tree grep on the candidate SHA found `pending brokered QA` only in negative test assertions, not in production prompt text. - The updated tests explicitly cover the engineer default skip wording and the `pull-request-and-merge` skipped-review path. - The issue comment posted on `2026-07-28T02:57:16Z` reports: focused wording checks passed, but repo-wide `ward exec test` failed in `cmd/ward/container_hostnet_test.go` (`TestBuildUpPlanTailnet`). - The PR comment posted on `2026-07-28T02:57:16Z` repeats the same blocker and says the branch cannot go green until that unrelated failure is cleared. - The live commit-status API for SHA `06b75094...` reports one surfaced check: `test / test (pull_request)` = `success` at `2026-07-28T02:57:00Z`. risks: - The surfaced PR status and the issue-thread narrative disagree: the visible PR check is green, but the recorded repo-wide `ward exec test` result is still failing. That mismatch makes acceptance ambiguous. - Because this QA pass is read-only and the full repo test run is not exposed as a passing check for this SHA, I cannot confirm the final acceptance criterion from live state alone. next steps: - Re-run or surface the full repo-declared validation for SHA `06b75094f92e4d54bbf1f08011f7da7929bc0c1f`, especially `ward exec test`. - If `TestBuildUpPlanTailnet` is still failing, treat this issue as waiting on that external blocker rather than on the wording change itself. - If the full suite is now green, update the issue/PR thread so the authoritative record matches the live check state. run_identity: KAI-DESKTOP-TOWER dispatcher framing: Read the issue title, body, and comment thread below as the QA brief. Inspect the candidate branch, any linked pull request, and the available checks in the live repository state. Return a structured QA verdict that a human can read at a glance. Do not edit files, commit, push, or otherwise change implementation state. </details>  — Codex, via `ward agent`

coilyco-ops commented

2026-07-28 03:05:09 +00:00

Author

Member

Exact-commit QA inspected PR #1578 at

coilyco-ops commented

2026-07-28 03:05:24 +00:00

Author

Member

Correction to the truncated preceding comment: exact-commit QA inspected PR #1578 at 06b75094f9. The wording and focused coverage look correct, and Forgejo Actions is green. QA remains blocked because full ward exec test reproduces the existing read-only-surface failures already tracked at #1565, including TestBuildUpPlanTailnet. This issue will take a fresh exact-commit QA pass after #1565 restores the repository-declared test gate. No merge is authorized before then.

Correction to the truncated preceding comment: exact-commit QA inspected PR #1578 at 06b75094f92e4d54bbf1f08011f7da7929bc0c1f. The wording and focused coverage look correct, and Forgejo Actions is green. QA remains blocked because full ward exec test reproduces the existing read-only-surface failures already tracked at #1565, including TestBuildUpPlanTailnet. This issue will take a fresh exact-commit QA pass after #1565 restores the repository-declared test gate. No merge is authorized before then.

coilysiren closed this issue

2026-07-28 03:07:18 +00:00

coilysiren referenced this issue from a commit

2026-07-28 03:07:20 +00:00

Update engineer QA skip wording (#1578)

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

coilyco-flight-deck/ward#1577

No description provided.

Rows
Columns