v1.60.2.0 feat: add complete ship handoffs

[fedster99]

Summary

• Make every /ship attempt end with a mandatory engineering handoff covering outcome, root cause, decisions, implementation, verification, operational risk, remaining work, and any user decision.
• Make ### Put simply the literal final section, with an extremely concise Why / What / How.
• Require faithful matched Before/After screenshots for user-visible UI changes before push; incomplete evidence blocks push and PR creation with no waiver.
• Propagate the contract through generated Claude, Codex, and Factory skills, coverage routing, deterministic regression tests, LLM behavior evals, and public documentation.
• Bump gstack to v1.60.2.0.

Motivation and root cause

/ship already automated landing mechanics, but its final user-facing output was underspecified. A run could stop at a PR URL, omit engineering context, or describe UI work without proving the visual change. The fix is an explicit, testable completion contract, with UI evidence moved into the pre-push verification gate.

Test Coverage

• Deterministic handoff regression coverage across the template plus generated Claude, Codex, and Factory variants.
• LLM instruction-quality judge and three-scenario behavior eval: non-UI success, UI success with matched evidence, and blocked UI evidence.
• Focused gate: 505 passed, 27 credential-gated evals skipped, 0 failed; 2,373 assertions across 11 files.
• Documentation gate: 336 passed, 0 failed; 1,283 assertions.
• Manual authenticated Sonnet behavior run passed all three scenarios; instruction-quality scores were clarity 5/5, completeness 5/5, actionability 4/5.
• The broad canonical suite was also attempted. Its failures are pre-existing environment issues in gbrain tests because their isolated test PATH cannot see jq; the same failures reproduce on an untouched origin/main worktree. Changed-path gates pass.

Pre-Landing Review

• Plan audit found one terminal-output contradiction; fixed by moving the PR URL into the engineering handoff and regression-locking the old phrase out.
• Adversarial review found three P2 issues: optional headings, prompt-only evaluation, and UI evidence occurring after publish actions. All three were fixed.
• No P0/P1 findings. Scope drift: clean. Generated-host parity: clean.

Design Review

No frontend or product UI files changed. Before/after screenshots are not applicable to this PR.

Eval Results

• New paid LLM evals are registered in touchfile selection and compile successfully.
• They skip locally when ANTHROPIC_API_KEY is absent.
• Equivalent authenticated CLI model evidence passed the instruction-quality and three-scenario behavior contracts.

Plan Completion

No repository plan file was detected. The requested engineering handoff, conditional UI proof, final concise explanation, host parity, tests, and documentation are complete.

Documentation

• README.md: updated the /ship example and skill catalog to show the full engineering handoff, concise Why / What / How, and matched UI proof gate.
• AGENTS.md: updated the release-skill reference with the same completion contract.
• docs/skills.md: added the detailed engineering handoff and UI evidence behavior to the /ship deep dive.

Coverage: /ship now has reference, example/how-to, and explanation coverage. No documentation debt or architecture diagram drift found.

Test plan

• Generate all host skill variants and verify byte-stable idempotency.
• Run deterministic handoff, host parity, coverage routing, validation, size-budget, and versioning tests.
• Run targeted documentation validation.
• Exercise non-UI, UI-complete, and UI-blocked handoffs with an authenticated model.
• Verify clean diffs and public-repo redaction scans.

🤖 Generated with Codex

[trunk-io]

Merging to main in this repository is managed by Trunk.

• To merge this pull request, check the box to the left or comment /trunk merge below.

After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here