Created
Jun 23, 02:43
Started
Jun 23, 12:18
Completed
Jun 23, 13:37
DevOps handoff
Type
Docs
Shape
—
Worktree Slug
document-agent-shell-test-gotchas
Repositories
mcritchie-studio
Release Train
—
Branch
feat/document-agent-shell-test-gotchas
Local URL
—
QA URL
—
Production URL
—
Acceptance Criteria
Expected Test Plan
No expected checks recorded.
Checks Run
No completed checks recorded.
Agent Context
Captures two test-run gotchas hit by real sessions: (1) non-interactive agent shells don't load direnv so ruby defaults to brew ruby@3.1/system 2.6 -> bundler RubyVersionMismatch; fix is PATH-prepend /opt/homebrew/opt/ruby@3.3/bin (apps pin 3.3.11). (2) the suite forks for parallelism which can segfault the pg gem on macOS (Ruby crash, not a test failure, intermittent); workaround PARALLEL_WORKERS=1 OBJC_DISABLE_INITIALIZE_FORK_SAFETY=YES. Surfaced by Avi during the PR #76 review.
Stage Timeline
Who handled each stage, the time it took (measured), and the model / tokens / cost reported (best-effort) — plus who's on it right now. — means the agent didn't report that metric.
Conversation
QA review feedback, agent handoffs, and follow-up notes for this task.
Doc-only change (only docs/agents/modules/testing.md; git status confirms). NOTE: bin/dor-check's auto-detect FALSE-FLAGS this as a code diff because the branch was cut from origin/release and the gate's committed-view base defaults to origin/main — so it swept in release-ahead-of-main commits (application_helper.rb, formatter.rb, etc.) that aren't mine. With the correct base it's clean: `DOR_CHECK_DIFF_BASE=origin/release bin/dor-check document-agent-shell-test-gotchas` → 'DoR n/a'. Filed as a bug (fix-dor-check-release-cut-base). The actual change here is doc-only and gate-exempt.
QA hold: PR #105 is a draft with empty Summary/Verification (stack showed /up 000 — nothing verified locally). Content in testing.md (agent-shell Ruby-PATH + pg fork-safety gotchas) is accurate and load-bearing. Fix: fill Summary/Verification, verify locally, mark Ready-for-review, then resubmit.
Resubmit after Avi's QA hold (process, not content — he confirmed the testing.md note is accurate + load-bearing). Fixed: filled PR #105 Summary + Verification, marked it Ready-for-review (was a draft from finish --pr). On 'verify locally / up 000': this is a doc-only change (kind: docs) with no app surface — the /up 000 was just an un-started worktree stack, not a regression; verification = git scope (only testing.md), dor-check DoR n/a (DOR_CHECK_DIFF_BASE=origin/release on the release-cut branch), and the two commands are the exact remedies hit live this session. Ready for re-review.
2-senior review (reviewer-select: alex-docs heavy + shannon light). alex-docs APPROVE — verified every fact against the live repo (.ruby-version 3.3.11, keg /opt/homebrew/opt/ruby@3.3/bin resolves to 3.3.11, pg arm64-darwin in Gemfile.lock); no fabrication, no doc traps. shannon APPROVE — accurate, paste-ready, well-placed. Both: optional non-blocking nit to hoist the callout to a shared subsection (applies to Turf too). Both criteria met = 2 approvals.
Sealed-bid sizing
Edit →Alex (PM)
—
Avi (PO)
—
Dev
—
Actual
—
We emailed a one-tap sign-in link to . It expires shortly and can only be used once.
No email? Check spam, or close this and try again.