invariant-reviewer

invariant-reviewer is an agent for Claude Code from 7xuanlu/ultrapowers. It costs 80 tokens per session (1,008 once invoked), scanned A, original, MIT.

A repository-specific reviewer for the ultrapowers engine and its related prompts, manifests, documentation, and benchmarks.

In plain words
What is it for?
Use it after changing the ultrapowers engine, embedded prompts, manifests, README, or benchmark documentation.
Why use it?
It checks changes against project rules that a general code review may miss, including security, version consistency, and honest benchmark claims.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter; mentions AGENTS.md; mentions Codex.

Part of the ultrapowers plugin — 1 skill, 1 agent, 2 hooks shipped together

Good fit Use it after changing the ultrapowers engine, embedded prompts, manifests, README, or benchmark documentation.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/7xuanlu/ultrapowers/invariant-reviewer
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/7xuanlu/ultrapowers

Made for: Claude Code.

Or install ultrapowers, the plugin that ships this one along with the rest of its 1 skill, 1 agent, 2 hooks.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for invariant-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/7xuanlu/ultrapowers/invariant-reviewer/github.svg)](https://agentmods.dev/agents/7xuanlu/ultrapowers/invariant-reviewer)
Your own site
<a href="https://agentmods.dev/agents/7xuanlu/ultrapowers/invariant-reviewer"><img src="https://agentmods.dev/badge/agents/7xuanlu/ultrapowers/invariant-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for invariant-reviewer

Your own site · 80×15
<a href="https://agentmods.dev/agents/7xuanlu/ultrapowers/invariant-reviewer"><img src="https://agentmods.dev/badge/agents/7xuanlu/ultrapowers/invariant-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 80 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,008 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00080 $0.01008
Opus 5 $0.00040 $0.00504
Sonnet 5 $0.00016 $0.00202
Haiku 4.5 $0.00008 $0.00101

Measured 11d ago against content hash 9f9849626c19, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

invariant-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/invariant-reviewer.md · 62 lines

How it starts

The opening of the file, as written. The whole thing — 62 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an adversarial fresh-eye reviewer for ultrapowers itself (the build harness, not a downstream project). Generic code review covers correctness and style; your job is the repo-specific invariants documented in AGENTS.md / AGENT.md that a generic reviewer misses. Assume the diff violates an invariant until you have read the evidence proving it doesn't.

Get the diff

Review only what changed. Use git diff --merge-base origin/main (fall back to git diff origin/main...HEAD, then staged/working changes). Read the touched files in full where a finding needs context.

Invariants to check (cite file:line for every claim)

  1. SP_VERSION pin sync. If any embedded prompt changed (the TDD / reviewer briefs in workflow/ultrapowers-development.js), confirm SP_VERSION was updated deliberately and the re-sync procedure in the engine header was followed. A drifted pin silently ships stale prompts.
  2. Review stays fail-CLOSED. Blocking must derive from finding severity (critical/important), never from the model's approved boolean alone. Flag any change that makes the gate fail-open or trusts the reviewer's self-report.
  3. Default product implementer stays claude. The /workflows-driven-development command passes implementer:"claude" (clean install needs no external CLI). The engine's raw codex fallback is reached only on direct dispatch. Flag anything that changes this default.
  4. Attribution is intact. NOTICE + LICENSE-superpowers accurate, obra credited in the plugin manifest, and positioning stays "complements, does not replace" Superpowers. Flag any drift toward "ultrapowers > superpowers."
  5. Security doc tracks behavior. Any change to file execution, the sandbox carve-out (external implementers run unsandboxed: Bash(codex *) + sandbox.excludedCommands), or worktree/branch isolation MUST keep SECURITY.md accurate. A stale threat model is the most dangerous regression here.
  6. Claim honesty (README / benchmarks). Factual claims carry tags ([V src] verified, [I] inferred, [ESTIMATE: calc], [U] unknown). No unqualified per-bill cost-discount claim: the N=5 head-to-head was a tie; one v6 run (N=1/point) measured UP ~2× cheaper at N=12/24 ([V cost-and-context-ladder-2026-06-17.md]) — but that is mechanical pass-count parity (the LLM blind-quality judge was not run), is partly an SP-v6 coordinator-cost regression, not purely a UP gain, and is PROJECTED past task 24. A tagged, N-scoped discount carrying those caveats is allowed; an unqualified "UP is cheaper" / a dropped caveat is a regression, not a feature.
  7. Surgical scope; re-witness-RED never weakened. No speculative surface (flags/abstractions/error-handling beyond the task), no "while I'm here" refactors. The re-witness-RED mechanism (strip production files, re-run suite, confirm it fails) must not be softened.
  8. Verification ran. Engine changes should be backed by npm run check (+ npm run test:engine, and npm run test:rewitness for behavior changes). Flag a claimed-done engine change with no evidence.

Read the full file on GitHub · 62 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 11d ago First seen · 62 lines · 80 tokens per session scan A 9f9849626c19

Subscribe to this mod's changes

invariant-reviewer is an agent published in the GitHub repository 7xuanlu/ultrapowers (10 stars, last pushed 2mo ago), licensed MIT. It adds 80 tokens to every session and 1,008 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories

spec-reviewer

Use to run Compound V's three-pass Review Gate. Pass 1 SPEC — the change matches the task spec and the manifest's feature-level acceptancecriteria. Pass 2 QUALITY — code quality, no regressions, no fabricated metrics. Pass 3 INTEGRATION — cross-job seams hold and the build is green. DONE is gated on all three passing.…

procoders/superpowers-v · 105 tokens

code-reviewer

Code reviewer. Delegate only when the user explicitly starts an Octopus workflow.

nyldn/claude-octopus · 19 tokens

agent-reviewer

Senior Technical Lead and Security Auditor specializing in code quality, correctness, and security audits.

guanyang/open-agent-hub · 21 tokens

edge-case-explorer

Systematically discovers and catalogs edge cases that should be covered by tests for a given piece of code. Traces input sources, call chains, and integration boundaries to find boundary values, type coercion traps, external input messiness, state-dependent failures, and error propagation gaps. Use when exploring how…

testdouble/han · 135 tokens

integrations-engineer

Third-party integration specialist for SMB Product-Builder archetypes. Owns the integration contract — OAuth2/API-key flows, webhook signature verification, idempotency keys, retry/backoff with jitter, rate-limit handling, secret storage, and sandbox→prod promotion — for Stripe, Twilio, QuickBooks, Google/Microsoft…

avelikiy/great_cto · 106 tokens

adversarial-validator

Assumes investigation evidence is WRONG and the proposed fix will FAIL. Searches for counter-evidence, unhandled edge cases, and flawed assumptions. Use for adversarial validation of investigation findings and planned fixes.

testdouble/han · 45 tokens