Adversarial Phoenix Liveview

Adversarial Phoenix LiveView Gate

A realtime-UI architecture review gate for Phoenix LiveView — pass/fail: a single blind reviewer subagent judges the work against this gate's rules with an adversarial mandate, and the work passes only when every rule is PASS or N/A. This skill renders verdicts; it never fixes the work.

The rules target one failure mode: LiveView written as a page-controller bolted to a JS framework, instead of what it is — a stateful process rendering change-tracked diffs over a lossy socket. Remounts that destroy state a patch would keep, broadcasts emitted where only one writer fires them, external calls that freeze the whole view, template constructs that silently disable change tracking, collections that grow per-socket memory without bound, events trusted because the button was hidden, and interactions designed against localhost latency. These all compile cleanly and demo perfectly; they surface as memory bloat, stale UIs, frozen views, and IDOR reports in production. Each rule carries an Evidence of violation paragraph so a reviewer can decide PASS/FAIL/N/A from artifact evidence alone. Sibling gates: adversarial-beam judges message-delivery semantics, adversarial-elixir judges paradigm fit — run all three for full coverage of substantial Phoenix work.

When to Apply

• A Phoenix LiveView feature — live views, live components, HEEx templates, PubSub-driven UI updates, presence, infinite scroll — is about to merge and needs an objective PASS/FAIL on its realtime-UI architecture.
• An agent (Claude, Codex) authored LiveView code and you want an independent check that it did not fake realtime correctness — controller habits, unscoped topics, unbounded assigns, unauthorized event handlers, zero-latency assumptions.
• A LiveView app is being prepared for real traffic and the localhost-era assumptions (no latency, one node, small data, trusted clients) need to be surfaced as verdicts.
• A production incident (memory bloat, frozen view, IDOR, stale UI) was fixed at one site and you want the same class hunted across the affected area.

Do not apply to non-Phoenix codebases (the reviewer prompt's precondition aborts with "GATE NOT APPLICABLE"), or when the user wants explanations and refactors rather than a verdict. Version note: every FAIL trigger rests on LiveView 1.0+ APIs (streams, async, JS commands); the missing-:key leg is version-gated to LiveView 1.1+ (colocated hooks never trigger a FAIL), and current_scope spellings are Phoenix 1.8 scaffolding — pre-1.8 targets are judged on the underlying shape (actor-filtered access). The reviewer prompt carries the applicability axes.

Review Protocol

Follow these steps exactly — the gate's value is that every review runs the same way.

1. Identify the target. Pin down exactly what is under review (a diff, a set of files, a PR) and note the ref/paths so the review runs against an unambiguous, fixed target. Record the stack facts (from mix.exs, the router, and config/) — phoenix/phoenix_live_view versions, scope module presence, PubSub/Presence usage, live_session blocks and their on_mount hooks, where global CSS lives. Include the repo root in the target description — several rules must search beyond the diff for context callers, router auth hooks, revocation flows, loading-class styling, and component definitions (the reviewer prompt lists them).
2. Load the rules. Read references/_sections.md and every rule file in references/ (all state-*.md, flow-*.md, async-*.md, render-*.md, stream-*.md, bound-*.md, trust-*.md, ui-*.md files).
3. Compose the reviewer prompt. Fill references/reviewer-prompt.md with the rules, the target, and the stack facts. The composed prompt must be fully self-contained — a reviewer sees no conversation history, so nothing may refer to context outside the prompt.
4. Dispatch one blind reviewer. Launch a single Task subagent whose entire input is the composed prompt — no conversation context, no commentary alongside it.
5. Render fail-closed. The reviewer's structured output is the verdict — there is no merge step. Overall verdict is PASS only when every rule is PASS or N/A; any single FAIL fails the gate. Never average, weigh severity, or waive a rule — a "minor" FAIL is a FAIL. If the reviewer returns "GATE NOT APPLICABLE" (not a Phoenix LiveView target), stop and report that instead of a verdict.
6. Render the verdict. Fill assets/templates/verdict.md. On FAIL, aggregate the reviewer's "missing for PASS" suggestions into the fix list, each with its location, ordered by category importance. Every rule whose final result is FAIL must appear in the fix list with a change concrete enough to apply as written — if the reviewer's suggestion only restates the violation, derive the fix from the rule's Correct example before rendering.

If the same rule flips verdicts across re-reviews of an unchanged target, or a human reads the evidence and overrides the verdict, that is a decidability bug in the rule — record it in gotchas.md and sharpen the rule; do not override the gate.

Verdict Format

The reviewer returns, per rule: PASS | FAIL | N/A, evidence (file:line or a quote — required for PASS as well as FAIL), and for every FAIL, the fix that flips the rule to PASS once applied — the named change plus its location, never a restatement of the violation. The final report follows assets/templates/verdict.md.

Rule Categories

Gotchas

Read gotchas.md before dispatching reviewers — it pre-records scope guards (sibling-gate boundaries, every-rule-fails-closed, mechanism-presence-not-appearance for ui-*, shapes-not-brands, the current_scope spelling note, the Repo-CRUD exemption, the wire-cost-not-breakage framing) so reviewers do not judge outside the rules.

Related Skills

• adversarial-beam — the runtime-semantics sibling gate: delivery guarantees, backpressure, supervision, shared-state races. This gate stops where the socket meets PubSub; that one owns what happens to the messages.
• adversarial-elixir — the paradigm-fit sibling gate: OO/enterprise habits ported onto the BEAM. Same protocol, complementary rules.
• staff-level-elixir — the advisory sibling: greenfield "which tool, which convention" guidance and judgment calls the gates deliberately exclude. Use it to write or fix code; use the gates to verdict it.

Reference Files