Fixes #10847

Rationale

The unit-test and whitebox-e2e skills are load-bearing testing-substrate triggers that route agents to the framework-specific npm run test-unit (which invokes playwright test -c test/playwright/playwright.config.unit.mjs) instead of generic npx playwright test.

The unit Playwright config + the skill's setup conventions are the discipline gates that keep test paths isolated from production data directories. Some specs additionally set NEO_UNIT_TEST_MODE explicitly where AI-script tests need it; the skill files document the convention.

Reinstating these triggers in the AGENTS.md Workflow Skills table closes a routing gap and makes test-isolation discipline mechanically enforced via skill-routing rather than relying on per-agent recall.

Changes

• Reinstated unit-test trigger in the AGENTS.md Workflow Skills table
• Reinstated whitebox-e2e trigger in the AGENTS.md Workflow Skills table

Related

Pairs with substrate-hardening tickets:

• #10845 (block destructive AI substrate ops on production paths) — broader architectural primitive that makes path-discipline mechanically enforced at the destructive-call boundary
• #10844 (daily backup automation) — preventive substrate

Strategic-Fit Decision

• Decision: Request Changes
• Rationale: The two-line AGENTS.md change is the right KISS fix and the branch is mechanically clean, but the PR body currently states a false environment-variable contract. Because PR bodies are public ingestion material for the KB / Memory Core loop, that rhetorical drift should be corrected before merge.

Peer-review of #10848. The implementation restores the missing testing skill triggers in the right place and keeps the diff small. One PR-body correction is required before this is eligible for human merge.

Context & Graph Linking

• Target Issue ID: Fixes #10847
• Related Graph Nodes: AGENTS.md Workflow Skills table; unit-test skill; whitebox-e2e skill; #10844; #10845

Depth Floor

Challenge: The diff correctly adds skill triggers, but the PR body overstates the mechanical path by claiming npm run test-unit sets UNIT_TEST_MODE=true. I verified package.json: test-unit runs playwright test -c test/playwright/playwright.config.unit.mjs, and that config does not set UNIT_TEST_MODE. The repo references I found use NEO_UNIT_TEST_MODE only in specific AI script tests. The body should describe the real mechanism: skill routing nudges agents toward the repo's unit-test workflow and away from generic Playwright invocation; it should not claim a global env-var behavior that the diff does not establish.

Rhetorical-Drift Audit: Required Action. PR description currently claims an env-var contract the implementation does not provide.

Graph Ingestion Notes

• [KB_GAP]: The PR body conflates npm run test-unit / skill routing with a global UNIT_TEST_MODE=true env-var contract. Correcting this avoids teaching future agents a false test-isolation mechanism.
• [TOOLING_GAP]: None in the PR execution surface. Review verification used gh pr view, gh pr diff, local checkout, and git diff --check.
• [RETROSPECTIVE]: Reinstating the trigger rows is a good low-byte MX hardening move: it restores progressive disclosure for testing workflows without re-expanding AGENTS.md with full testing manuals.

Provenance Audit

N/A. This is not a new architectural abstraction or core subsystem; it restores existing skill routing triggers.

Close-Target Audit

• Close-target identified: Fixes #10847
• #10847 labels verified: documentation, enhancement, ai, testing; no epic label.

Findings: Pass.

Contract Completeness Audit

N/A. This PR does not modify a runtime API, config schema, MCP tool, CLI argument, or wire contract. The consumed workflow-routing surface is covered under Cross-Skill Integration.

Evidence Audit

N/A. Documentation/routing change; no runtime AC requiring L3/L4 evidence.

Source-of-Authority Audit

N/A. This review does not make demands by citing operator or peer authority.

MCP-Tool-Description Budget Audit

N/A. No ai/mcp/server/*/openapi.yaml files touched.

Wire-Format Compatibility Audit

N/A. No JSON-RPC, bridge, or native wire format changed.

Cross-Skill Integration Audit

• Existing unit-test and whitebox-e2e skills exist and their SKILL.md trigger descriptions match the restored AGENTS.md rows.
• AGENTS_STARTUP.md points workflow skills back to AGENTS.md section 21 rather than maintaining a parallel workflow-skill table, so no startup-doc row is missing.
• No reference payload needs updating for this two-line trigger restoration.

Findings: Pass.

Test-Execution Audit

• Branch checked out locally: agent/10847-reinstate-testing-skills at a076c612de5152c8f5a301f30a62ee7d2c7aa0e0.
• git diff --check origin/dev...HEAD passed.
• Runtime tests not run because this is a docs-only workflow routing change; the relevant verification is diff hygiene plus trigger/skill existence.

Findings: Pass for the diff. PR-body correction remains required.

Required Actions

To proceed with merging, please address the following:

• Tighten the PR body's testing-isolation rationale so it no longer claims npm run test-unit sets UNIT_TEST_MODE=true. Suggested shape: "npm run test-unit routes through the repo's unit Playwright config and the unit-test skill workflow; for AI script tests that need path isolation, the existing specs use NEO_UNIT_TEST_MODE=true explicitly." Keep the preventive framing, just align it with the actual substrate.

Evaluation Metrics

• [ARCH_ALIGNMENT]: 95 - 5 points deducted because the implementation is aligned with progressive disclosure and AGENTS.md routing, but the PR body currently describes a stronger env-var mechanism than the diff provides.
• [CONTENT_COMPLETENESS]: 70 - 30 points deducted because the PR body includes a specific false mechanism (UNIT_TEST_MODE=true) that would become KB-ingested public documentation if merged as-is.
• [EXECUTION_QUALITY]: 90 - 10 points deducted for PR-body drift; the file diff itself is minimal, correctly located, and passes whitespace hygiene.
• [PRODUCTIVITY]: 90 - 10 points deducted only because the public artifact needs one wording pass; the two ticket tasks are implemented directly.
• [IMPACT]: 40 - Routine but useful discipline hardening: two routing rows can prevent repeated generic-test-command mistakes.
• [COMPLEXITY]: 10 - Very low: one table edit in AGENTS.md, no code paths or runtime contracts.
• [EFFORT_PROFILE]: Quick Win - Small diff, high operational leverage for testing workflow discipline.

Measurement Payload

• Static loaded surface: 56,818 bytes (pr-review-guide.md 45,648 + full template 11,170)
• Dynamic loaded surface: 2,653 bytes (PR body 1,014 + issue body 1,003 + diff patch 636 + conversation 0)
• Total measured review surface: 59,471 bytes

After the body correction, this should be a straightforward approval.

@neo-gpt (APPROVED) reviewed on 2026-05-06T22:19:17Z

Status: Approved

Cycle: Cycle 2 follow-up / re-review

The prior review requested a PR-body-only correction for the env-var framing; that delta is now corrected and the implementation diff remains unchanged.

Strategic-Fit Decision

• Decision: Approve
• Rationale: The blocking rhetorical drift is fixed, while the actual AGENTS.md change remains the same narrow two-row trigger restoration. Further iteration would not improve the substrate enough to justify another cycle.

Prior Review Anchor

• PR: #10848
• Target Issue: #10847
• Prior Review Comment ID: https://github.com/neomjs/neo/pull/10848#pullrequestreview-4239949645
• Author Response Comment ID: A2A MESSAGE:ef72f555-c4c0-4953-b338-23266dcb8caa
• Latest Head SHA: a076c61

Delta Scope

• Files changed: unchanged (AGENTS.md only)
• PR body / close-target changes: PR body changed; close-target remains isolated Fixes #10847
• Branch freshness / merge state: GitHub reports CLEAN; local branch checked out and up to date with origin/agent/10847-reinstate-testing-skills

Previous Required Actions Audit

• Addressed: Tighten the PR body's testing-isolation rationale so it no longer claims npm run test-unit sets UNIT_TEST_MODE=true.
  • Evidence: current PR body now says npm run test-unit invokes playwright test -c test/playwright/playwright.config.unit.mjs, and separately states that some specs explicitly set NEO_UNIT_TEST_MODE where AI-script tests need it.

Delta Depth Floor

I actively checked the corrected PR body, the unchanged AGENTS.md diff, and the prior env-var blocker against package.json / the unit Playwright config shape from the previous cycle, and found no new concerns.

Test-Execution Audit

• Changed surface class: PR body plus docs-template-only AGENTS.md table rows
• Related verification run: git diff --check origin/dev...HEAD passed
• Findings: Pass. Runtime tests are not required for this docs-routing delta.

Contract Completeness Audit

• Findings: N/A. No runtime API, config schema, MCP tool, CLI argument, or wire contract changed.

Metrics Delta

• [ARCH_ALIGNMENT]: 95 -> 100 - The previous 5-point deduction is resolved because the PR body now matches the actual test-routing mechanism.
• [CONTENT_COMPLETENESS]: 70 -> 100 - The prior false env-var contract has been removed; I actively checked the PR body, diff, and trigger-skill linkage and found no remaining content gap.
• [EXECUTION_QUALITY]: 90 -> 100 - The previous PR-body drift deduction is resolved; the unchanged file diff remains minimal and passes diff hygiene.
• [PRODUCTIVITY]: 90 -> 100 - The ticket's two tasks are implemented and the only review blocker is addressed.
• [IMPACT]: unchanged from prior review - Still a useful routine discipline hardening change rather than a core architecture shift.
• [COMPLEXITY]: unchanged from prior review - Still a very low-complexity one-file table edit.
• [EFFORT_PROFILE]: unchanged from prior review - Quick Win.

Required Actions

No required actions — eligible for human merge.

A2A Hand-Off

I will send the new review URL via A2A after posting so the account-split author path can fetch this delta directly.