Repository navigation
Description
Problem
The three suites in tests/skills/source/lib/ (lib.bats, gates.bats, masked-status.bats, about 1,150 lines of code between them) test shell behavior in packages/agents/source/lib/lib.sh, agent.sh, signing.sh and gates.py. Much of that behavior already has a Go owner on main (internal/config, agentrepo config) or is covered by a Go test, so the Bats cases are duplicate coverage that slows the suite and has to be edited by every later port. The cleanup does not depend on the N-2 window, because it deletes tests, not the shell or its fallback, so it merges first.
Part of #588 and #321. Deletes test cases only; no shell under packages/agents/source/lib/ changes.
Acceptance criteria
- AC1: Given every case in the three suites, when the retirement ledger is read, then each case is marked deleted with the Go test that covers it, kept with the later child that will retire it, or kept with the reason it needs a real shell, and no case is unaccounted for.
- AC2: Given the merged change, then every case marked deleted is absent from the suites, and every kept case still passes.
- AC3: Given the shell under
packages/agents/source/lib/and the generated skills, then neither changes, andmise run check-generated-skillspasses. - AC4: Given the ledger, when it is read, then each deleted case is marked as covered by a native Go test or by a shell-integration Go test (one that drives the generated wrappers), and each covering test is marked with the platforms on which it executes and not skips.
- AC5: Given the Go tests named in the ledger, when CI runs on the merged head, then each executes (is not skipped) on every platform the ledger promises for it, among Linux, macOS and Windows, and passes.
- AC6: Non-functional: the pull request reports the median wall time of three Bats runs of the three suites before and after the change, on the same runner, as a measurement with no pass threshold, and no coverage of still-shipped shell behavior is lost.
Scope and non-goals
Scope: the ledger and the deletion of duplicate cases in the three suites.
Non-goal: deleting any shell, the capability probe or the fallback; child 10 owns the area's non-owner fallbacks after the N-2 window, and the cutover issues under #588 delete the protected ones. Non-goal: porting new behavior to Go. Out of scope: suites outside tests/skills/source/lib/.
Approach
Meet AC1 to AC6. Build the ledger from the case names first, delete only cases whose Go coverage already exists on main, and leave the rest for the change that ports them. The mechanism is the implementer's choice within those outcomes. If the deleted cases would exceed the size limit, the remainder goes to a follow-up issue filed from this one.
Dependencies
Blocked by: none, for the cases whose Go coverage is on main.
Prerequisite: the Go tests for internal/config and agentrepo config are on main. Child 08 and child 09 extend this child's ledger.
The pull request stays below 800 lines of code and 2,000 prose lines; excess goes to a follow-up issue, not to a later retirement child.
Size
S. One pull request, deletions only, at most 800 lines of code.
Verify plan
- AC1: a script or test lists the case names in the three suites and checks each against the ledger.
- AC2:
git diffshows only deletions of ledger-marked cases, and the Bats run for the suites passes. - AC3:
git diff --statshows no file outsidetests/skills/source/lib/and the ledger, andmise run check-generated-skillspasses. - AC4: a check that every deleted ledger row names its covering test, its coverage kind and its executing platforms.
- AC5: the CI job logs for the named Go tests on each promised platform, showing the tests ran and were not skipped.
- AC6: the three-run median Bats timings before and after are recorded in the pull request, with the coverage diff of the ledger.
- Definition of done: D1 each AC is verified at the merged head, with the commands and results cited in the pull request's Verification. D2 an independent review approves the exact merged head, and every finding has a recorded disposition. D3 required checks pass with no gate loosened. D4 the leak scans the repository specifies are clean, and every post made for this work passes the per-post leak review before it is posted. D5 every commit is signed. D6 the documentation matches the change or the Verification says why there is no impact. D7 every commit has an
Assisted-bytrailer and the pull request has theprovenance:agent-assistedlabel. D8 the pull request stays within the size limits. D9 follow-ups are filed as issues and listed under Related. D10 the merge closes the issue through a closing reference, the board shows Done, and the finalStatus:comment names the merge commit.
Risks
Deleting a case whose Go equivalent is weaker loses coverage. The ledger names the covering test for each deletion, and the review checks it. Undo: revert the change, which restores the cases.
Operator authority
None.
Metadata
Metadata
Assignees
Labels
Type
Fields
Priority
Work type
Projects
- StatusShow more project fieldsBacklog
Activity
tbhb-agent commentedon Oct 27, 2026
Refinement: READY — The shared library Bats suites are retired where Go already covers them
Model: gpt-6-astra
Child 01 The shared library Bats suites are retired where Go already covers them
Verdict: READY
Checked: Prior coverage findings remain closed; refresh the acknowledged child references when filing. Gaps: None. Split: None.
Definition of done
The definition of done for this issue (Chore). Tick each item and give its evidence.
check-leaks.shwith the repository's private-strings list, and the leak review of each post.docs/, CHANGELOG or help text), or the Verification section says "No doc impact" and why.provenance:agent-assistedlabel.Status:comment names the merge commit.mise run checkis green, and so are the coverage floorsmise run cover-go-checkandmise run cover-py-check, whichcheckdoesn't run.packages/agents/source/commits whatmise run generate-skills, thenmise run skill-tokens, thenapm installrewrite, in that order, andcheck-generated-skills(part oflint) passes.docs/agent-skills-migration.mdor the adoption docs, or the PR's Verification says "No consumer impact" and why.Round: 1 of 3