agentleFS
iktok90-designiktok90-design/ai-smart-contract-auditor/skills/oracle-manipulation/SKILL.md

oracle-manipulation

Skill36 starsChanged 5 months agoMITView on GitHub

Detect oracle manipulation risks — spot-price reads from AMMs, stale Chainlink answers, single-source dependence, TWAP gaming. Activate whenever code reads a price, conversion rate, exchange rate, or `getReserves`, `latestAnswer`, `latestRoundData`, `consult`, `quote`, `slot0`, `observe`, `getAmountsOut`.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Only one oracle, no fallback, no cross-check against a second source.

    line 45
  2. Very specific:

    Use observe() for TWAP with sufficient lookback (≥30min for shallow pools).

    line 51
  3. Distinctive:

    Use multiple oracles with deviation check (e.g.

    line 86
  4. Very specific:

    Avoid slot0 and spot-price reads for anything that affects user funds.

    line 89

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Oracle manipulation detection

When this applies

Any code that consumes a price feed or derives a price from on-chain state:

  • Chainlink: latestAnswer, latestRoundData, getRoundData
  • Uniswap V2: getReserves, getAmountsOut, price0Average
  • Uniswap V3/V4: slot0, observe, consult
  • Curve: get_virtual_price, get_dy
  • Balancer: getRate
  • Custom oracle modules, on-chain TWAPs
  • LP token pricing, collateral valuation, liquidation thresholds, mint/burn rates

Detection patterns

Spot price from AMM (CRITICAL — funds at risk)

(uint112 r0, uint112 r1,) = pair.getReserves();
uint256 price = r1 * 1e18 / r0;       // ← flash-loan manipulable in one block

Stale Chainlink data (HIGH)

(, int256 answer,,,) = feed.latestRoundData();   // ← ignores updatedAt and answeredInRound

Required checks:

  • updatedAt != 0
  • block.timestamp - updatedAt <= heartbeat (with chain-specific tolerance)
  • answeredInRound >= roundId
  • answer > 0
  • Heartbeat sanity: ETH/USD is 1h on mainnet but 24h on some L2s — verify

latestAnswer (deprecated, no freshness) (HIGH)

int256 p = feed.latestAnswer();   // ← deprecated, prefer latestRoundData

Single-source dependence (HIGH)

Only one oracle, no fallback, no cross-check against a second source.

Uniswap V3 slot0 without TWAP (CRITICAL)

(uint160 sqrtPriceX96,,,,,,) = pool.slot0();   // ← spot, manipulable

Use observe() for TWAP with sufficient lookback (≥30min for shallow pools).

Insufficient TWAP window (HIGH)

TWAP < 5 min on volatile assets or low-liquidity pools is gameable across two blocks.

Read-only reentrancy on oracle (HIGH)

Reading getVirtualPrice / getReserves mid-callback — see [[reentrancy]] §Read-only.

LP token priced via reserves directly (CRITICAL)

LP token price should use the fair LP price formula (Alpha Homora's method), not reserve / totalSupply.

Price scaling errors (HIGH)

Chainlink answers come in feed-specific decimals; mixing ETH/USD (8 decimals) with token (18 decimals) without normalizing produces silent off-by-1e10 errors.

Sequencer uptime ignored (HIGH on L2s)

On Arbitrum, Optimism, Base — must check SequencerUptimeFeed to avoid stale prices when sequencer is down.

Severity rubric

PatternSeverity
Spot-price from AMM used in liquidation/mintCritical
Uniswap V3 slot0 used for collateral pricingCritical
LP token priced via raw reserve / totalSupplyCritical
latestRoundData without freshness/round checksHigh
Single oracle, no fallback, financial-critical pathHigh
Sequencer uptime check missing on L2High
Insufficient TWAP window (<5min on shallow pool)High
Stale heartbeat for known-volatile assetHigh
latestAnswer (deprecated) usedMedium
Decimal-scaling mistakeHigh

Remediation patterns

  • For Chainlink: validate (roundId, answer, , updatedAt, answeredInRound) — guard against stale + zero + non-positive + round-mismatch.
  • Use multiple oracles with deviation check (e.g. Chainlink + Uniswap V3 TWAP with ≥30min window).
  • For LP tokens, use fair-LP-price math: 2 * sqrt(r0 * r1 * p0 * p1) / totalSupply.
  • On L2s, integrate Chainlink's SequencerUptimeFeed (Arbitrum, Optimism, Base).
  • Avoid slot0 and spot-price reads for anything that affects user funds.

False-positive notes

  • Display-only prices (UI/event metadata, not affecting fund flow) — Info-only.
  • getReserves used purely for ratio diagnostics with no downstream effect — Info.

Related

  • [[flash-loan-attacks]] — most oracle attacks are funded by flash loans
  • [[reentrancy]] — read-only reentrancy is an oracle-staleness issue
  • [[mev-frontrunning]] — TWAP-window timing

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.