uinaf/autoreview
Runs one independent review of completed local changes, pull requests, branch diffs, or commits through the installed autoreview Go CLI using Codex, Claude, Cursor, or Grok; gathers authoritative acceptance criteria, selects or honors one provider, validates findings, applies scoped fixes, verifies and reruns accepted fixes, and safely reports reproducible CLI defects. Use when the user asks for an autoreview, a review of their code, pull request, or changes, an automated PR or second-model code review, a final tool-backed review, or review closeout after builder verification. Do not use as builder verification or a multi-reviewer panel.
—
—
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
View guide
Low
Low-risk findings worth noting
providers.mdreferences/
Provider selection
Select one provider for the entire review. Honor the user's provider, model, and reasoning-effort choices first, then trusted repository or account configuration. Pass explicit user model and effort choices as CLI flags. If no source chooses a provider, select one installed harness that satisfies the task and explain the choice. Do not run multiple providers, fall back after failure, or claim consensus.
| Engine | Default model | Strict authentication | Important constraint |
|---|---|---|---|
codex | gpt-5.6-sol | CODEX_API_KEY or OPENAI_API_KEY | Native mode uses normal session-backed login |
claude | claude-opus-5 | ANTHROPIC_API_KEY | Effort supports low, medium, high, xhigh, or max |
cursor | cursor-grok-4.5-high-fast | CURSOR_API_KEY | Explicit CLI selection implies web when unset; effort is encoded in the model ID |
grok | grok-4.5 | XAI_API_KEY | Native mode uses grok login; tools, memory, plans, and subagents stay disabled |
For Cursor, pass a requested compatible model with --model and never add
--reasoning-effort. If the user requests Cursor plus a separate effort value
without a compatible model ID, report the unsupported combination instead of
guessing a model.
All adapters capability-probe the installed executable and use one explicit model with no model fallback. A capability or authentication failure is an operational result, not permission to switch engines.
Strict mode runs with empty provider state and a constrained environment. Native mode preserves the provider's normal environment and login but still runs the review in an empty temporary workspace with only the frozen bundle. Native is the default; select strict explicitly when the stronger provider-state boundary and a supported API key are required.
Web access defaults off. Codex enables its search surface only when authorized; Claude exposes only WebSearch; Grok exposes only WebSearch and WebFetch. Cursor has no documented per-run web disable, so explicit CLI selection makes an otherwise-unset value true; configured engine selection does not. Explicit false still fails capability preflight.