aif-review (skill-context override)

Status: shipped-beta (AIF-native skill-context override) · Ships to: npm lane — copied into the consumer's .ai-factory/skill-context/aif-review/SKILL.md by the SHIPPED_DOCS-derived §3c copy loop (setup.d/20-agents.sh:74) · Fires at: mandatory-read by AIF's own background sidecars (setup.d/20-agents.sh:56) — concretely, AIF's aif-review skill and its background review-sidecar at review time

Covers census rows C22 and E16 — the same file rows in both families.

What it is

Skill-context override at packages/core/templates/shared/skill-context/aif-review/SKILL.md. File name aif-review/SKILL.md; frontmatter name: aif-review-project-context — the name E16's census row quotes. H1: "aif-review skill-context — anti-tautology two-AI review conventions". It carries "project-specific review conventions injected into AI Factory's aif-review skill (and its background review-sidecar)" — the anti-tautology / two-AI test-review checks this project requires in addition to AIF's generic review (…/aif-review/SKILL.md:7). The generic correctness/security/performance review stays owned by AIF's aif-review SKILL.md; this file augments it (…/aif-review/SKILL.md:8). The delivery route is deliberate: rather than ship a colliding review-sidecar agent, "aif-review gets our anti-tautology test-review content" (setup.d/20-agents.sh:58).

How it works

• Dual-implementation status. <!-- @dual-pair: review-sidecar --> + <!-- spec-of: agents/review-sidecar.md --> (…/aif-review/SKILL.md:10-11). Census C10 is agents/review-sidecar.md, the portable SSOT: the marker comment says "Portable SSOT for this content is agents/review-sidecar.md. This file is the AIF-native delivery channel (skill-context override) for the same anti-tautology spec … Edit the SSOT first, then mirror here." (…/aif-review/SKILL.md:12-14).
• Premise. The highest-value review signal is test quality, not just code correctness; apply the conventions as project-level overrides — when they add a check AIF's defaults lack, perform it and include it in the output (…/aif-review/SKILL.md:16).
• Stance. Review the diff as a cold external reviewer — read the changed code as if you had never seen it and did not write it; be skeptical of comments and commit messages; read the actual diff (…/aif-review/SKILL.md:18).
• The seven MUST-checks. Tautological tests (assertions true by construction — "if I removed this assertion, what bug could now ship?"), mock-only tests, missing edge cases, test name ≠ behaviour, test independence, React/Next anti-patterns, React testing anti-patterns (…/aif-review/SKILL.md:22-36).
• Output augmentation. A ### Test-Quality Review section listing each finding with severity (BLOCKER/MAJOR/MINOR), file:line, "what I saw", "why it's a problem", and a one-line fix; report only — never modify code (…/aif-review/SKILL.md:38-40).

Satellites & companions

• ADAPTS: agents/review-sidecar.md (census C10) — same anti-tautology spec, different delivery channel; the AIF-native skill-context override instead of a colliding agent file.
• USES: AIF's aif-review skill + review-sidecar as the host pipeline whose output this file augments (the mandatory-read wiring, setup.d/20-agents.sh:56-58).
• ADDS: the two-AI rationale — the implementer wrote code and tests in one head; the reviewer's value is being a different head (…/aif-review/SKILL.md:20).
• Census satellites: B35, C10.

Anchors

• packages/core/templates/shared/skill-context/aif-review/SKILL.md:2 — name: aif-review-project-context
• packages/core/templates/shared/skill-context/aif-review/SKILL.md:5 — # aif-review skill-context — anti-tautology two-AI review conventions
• packages/core/templates/shared/skill-context/aif-review/SKILL.md:7 — > **Authoritative for:** project-specific review conventions injected into AI Factory's `aif-review` skill (and its background `review-sidecar`) …
• packages/core/templates/shared/skill-context/aif-review/SKILL.md:10 — <!-- @dual-pair: review-sidecar -->
• packages/core/templates/shared/skill-context/aif-review/SKILL.md:11 — <!-- spec-of: agents/review-sidecar.md -->
• packages/core/templates/shared/skill-context/aif-review/SKILL.md:12 — <!-- Portable SSOT for this content is agents/review-sidecar.md. This file is the AIF-native
• packages/core/templates/shared/skill-context/aif-review/SKILL.md:18 — ## Review the diff as a cold external reviewer
• packages/core/templates/shared/skill-context/aif-review/SKILL.md:22 — ## MUST additionally check — test quality
• packages/core/templates/shared/skill-context/aif-review/SKILL.md:38 — ## Output augmentation
• setup.d/20-agents.sh:56 — # AIF's own background sidecars MANDATORY-read .ai-factory/skill-context/<skill>/SKILL.md
• setup.d/20-agents.sh:58 — # instead of shipping colliding agents: aif-review gets our anti-tautology test-review content;

All anchors at framework pin b069c59328aa8e08671d3a19fd25efa618dfc1c5 (fetched 2026-09-11 from artyhoo/getff staging; ancestry of census pin a1337cb301 verified).