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.
npx skills add mtarcure/claude-vibe-squad --skill defi-invariant-checkgit clone --depth 1 https://github.com/mtarcure/claude-vibe-squadWrote 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.
[](https://agentmods.dev/skills/mtarcure/claude-vibe-squad/defi-invariant-check)<a href="https://agentmods.dev/skills/mtarcure/claude-vibe-squad/defi-invariant-check"><img src="https://agentmods.dev/badge/skills/mtarcure/claude-vibe-squad/defi-invariant-check/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.
<a href="https://agentmods.dev/skills/mtarcure/claude-vibe-squad/defi-invariant-check"><img src="https://agentmods.dev/badge/skills/mtarcure/claude-vibe-squad/defi-invariant-check.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00072 | $0.02276 |
| Opus 5.5 | $0.00029 | $0.00910 |
| Sonnet 5.5 | $0.00014 | $0.00455 |
| Haiku 4.5 | $0.00007 | $0.00228 |
Grade A, and why
defi-invariant-check 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 27d 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.
How it starts
The opening of the file, as written. The whole thing — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.
defi-invariant-check
DeFi-specific invariant authoring and testing discipline. Use this as a sub-skill during
evm-audit-flow when the target is a DeFi protocol (AMM, lending, yield, bridge, stablecoin,
perps). Generic Echidna/Medusa campaigns miss DeFi-specific invariants — this skill provides
the property checklist. All tools are the native host CLIs (echidna, medusa, forge,
halmos) run in the target project directory; there is no wrapper layer.
Invariant categories
Token accounting invariants (always required):
totalSupply == sum(balanceOf(all_holders))— no spurious minting or burning.reserve0 * reserve1 >= k(for constant-product AMMs) — k-invariant never decreases except via fee accrual.totalBorrow <= totalDepositfor all lending pools — protocol cannot lend out more than deposited.sum(userShares) == totalSharesfor yield vaults — share accounting is consistent.
Access-control invariants:
- Only authorized roles can set oracle addresses, fee parameters, or pause state.
onlyOwner/onlyGovernancefunctions cannot be called by arbitrary external accounts.- Upgradeable proxy implementation slots cannot be modified except through the upgrade pathway.
Economic/oracle invariants:
- Flash loan callbacks cannot leave the protocol in a net-loss state after the loan repayment.
- Liquidation proceeds must cover the borrowed amount plus protocol fee — no bad-debt accrual from liquidations.
2026 incident-derived oracle-input robustness invariants (highest-priority economic class):
- Spot versus TWAP: a price used for collateral, borrowing, liquidation, minting, or
redemption must reject or clamp a same-block spot price when
abs(spot - TWAP) / TWAP > maxDeviation; no value-affecting path may silently fall back to the divergent spot. This targets the oracle/valuation-manipulation class represented by Moonwell (2026-08-27), Lazy Summer (2026-07-06), Makina (2026-01-20), and Edel (2026-07-01) inV114-51. - Low-liquidity collateral: within the protocol's configured manipulation budget, an
atomic trade against a thin market cannot increase an account's borrowable value or
liquidation proceeds by more than the attacker's net cost plus the configured safety
margin. Reject collateral whose observable depth cannot support the valuation notional.
This is the Moonwell low-liquidity-collateral oracle shape from
V114-51. - NAV manipulation bound: absent external profit or loss, a deposit/reprice/redeem round
trip across any supported asset must not return more value than entered after fees and
rounding tolerance, and one manipulable component price must not move total NAV beyond
its configured exposure and deviation caps. This targets the Lazy Summer, Makina, and
Edel valuation/NAV incident class from
V114-51.
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.
- 27d ago First seen · 158 lines · 72 tokens per session scan A e9bf84a2a33c
defi-invariant-check is a skill published in the GitHub repository mtarcure/claude-vibe-squad (165 stars, last pushed 25d ago), licensed MIT. It adds 72 tokens to every session and 2,276 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-09-19.
Other skills, from other repositories
money
Skill "money" from Hmbown/Codewhale, covering money, when to use, setup, workflow and non-goals.
dispute-respond
Prepare and explicitly file an evidence-bound payment dispute response using native receipt proof, approval, provider execution, and independent readback.
purchase-approval
Decide whether one proposed purchase is allowed under an explicit procurement policy before any spend flow starts, refusing budget overages and unapproved vendors instead of guessing.
pay
Pay one external payment requirement within exact parent authority, through a payment method the payer holds, and read the paid resource back.
charge
Quote, then verify and settle, one buyer's payment for a resource you sell, with approval and provider readback.
refund
Refund all or part of one settled payment on its original rail, with approval, idempotency, and readback.