AgentSkillsCN

web3-foundry

编写基于Foundry的测试与脚本。触发短语包括:“Foundry测试”、“编写测试”、“模糊测试”、“分叉测试”、“不变量测试”、“部署脚本”、“Gas基准测试”、“覆盖率测试”,或在测试/脚本/目录下进行相关工作时使用。

SKILL.md
--- frontmatter
name: web3-foundry
description: Write Foundry-based tests and scripts. Trigger phrases - foundry testing, write test, fuzz test, fork test, invariant test, deploy script, gas benchmark, coverage, or when working in tests/ or scripts/ directories.

Foundry Testing & Script Skill

Rules and patterns for Foundry tests. Find examples in the actual codebase.

Bundled References

ReferenceContentWhen to Read
./references/test-infrastructure.mdConstants, defaults, mocksWhen setting up tests
./references/cheat-codes.mdCommon cheatcode patternsWhen using vm cheatcodes
./references/invariant-patterns.mdHandlers, stores, invariantsWhen writing invariant tests
./references/formal-verification.mdHalmos, Certora, symbolic execWhen proving correctness
./references/deployment-scripts.mdScript patterns, verificationWhen writing deploy scripts
./references/deployment-checklist.mdPre-mainnet deployment stepsBefore deploying to production
./references/gas-benchmarking.mdSnapshot, profiling, CIWhen measuring gas performance
./references/sablier-conventions.mdSablier-specific patternsWhen working in Sablier repos

Test Types

TypeDirectoryNamingPurpose
Integrationtests/integration/concrete/*.t.solBTT-based concrete tests
Fuzztests/integration/fuzz/*.t.solProperty-based testing
Forktests/fork/*.t.solMainnet state testing
Invarianttests/invariant/Invariant*.t.solStateful protocol properties
Scriptsscripts/solidity/*.s.solDeployment/initialization

1. Integration Tests (Concrete)

Naming Convention

PatternUsage
test_RevertWhen_{Condition}Revert on input
test_RevertGiven_{State}Revert on state
test_When_{Condition}Success path

Rules

  1. Stack modifiers to document BTT path (modifiers are often empty - just document the path)
  2. Expect events BEFORE action - vm.expectEmit() then call function
  3. Assert state AFTER action - Check state changes after function executes
  4. Use revert helpers for common patterns (expectRevert_DelegateCall, expectRevert_Null)
  5. Named parameters in assertions - assertEq(actual, expected, "description")

Mock Rules

  1. Place all mocks in tests/mocks/
  2. One mock per scenario (not one mega-mock)
  3. Naming: *Good, *Reverting, *InvalidSelector, *Reentrant

2. Fuzz Tests

Naming Convention

testFuzz_{FunctionName}_{Scenario}

Rules

  1. Bound before assume - _bound() is more efficient than vm.assume()
  2. Bound in dependency order - Independent params first, then dependent
  3. Never hardcode params with validation constraints
  4. Document fuzzed scenarios in NatSpec

Bounding Pattern

solidity
// 1. Bound independent params first
cliffDuration = boundUint40(cliffDuration, 0, MAX - 1);

// 2. Bound dependent params based on constraints
totalDuration = boundUint40(totalDuration, cliffDuration + 1, MAX);

3. Fork Tests

Rules

  1. Create fork with vm.createSelectFork("ethereum")
  2. Use deal() to give tokens to test users
  3. Use assumeNoBlacklisted() for USDC/USDT
  4. Use forceApprove() for non-standard tokens (USDT)

Token Quirks

TokenIssueSolution
USDC/USDTBlacklistassumeNoBlacklisted()
USDTNon-standardforceApprove()
Fee-on-transferBalance diffCheck actual received amount

4. Invariant Tests

Architecture

code
tests/invariant/
├── handlers/     # State manipulation (call functions with bounded params)
├── stores/       # State tracking (record totals, IDs)
└── Invariant.t.sol

Rules

  1. Target handlers only - targetContract(address(handler))
  2. Exclude protocol contracts - excludeSender(address(vault))
  3. Use stores to track totals for invariant assertions
  4. Early return in handlers if preconditions not met

5. Solidity Scripts

Rules

  1. Inherit from BaseScript with broadcast modifier
  2. Use env vars: ETH_FROM, MNEMONIC
  3. Simulation first, then broadcast

Commands

bash
# Simulation
forge script scripts/Deploy.s.sol --sig "run(...)" ARGS --rpc-url $RPC

# Broadcast
forge script scripts/Deploy.s.sol --sig "run(...)" ARGS --rpc-url $RPC --broadcast --verify

Running Tests

bash
# By type
forge test --match-path "tests/integration/concrete/**"
forge test --match-path "tests/fork/**"
forge test --match-contract Invariant_Test

# Specific test
forge test --match-test test_WhenCallerRecipient -vvvv

# Fuzz with more runs
forge test --match-test testFuzz_ --fuzz-runs 1000

# Coverage
forge coverage --report lcov

Debugging

Verbosity Levels

FlagShows
-vLogs for failing tests
-vvLogs for all tests
-vvvStack traces for failures
-vvvvStack traces + setup traces
-vvvvvFull execution traces

Console Logging

solidity
import { console2 } from "forge-std/console2.sol";

console2.log("value:", someValue);
console2.log("address:", someAddress);
console2.logBytes32(someBytes32);

Debugging Commands

bash
# Trace specific failing test
forge test --match-test test_MyTest -vvvv

# Gas report for a test
forge test --match-test test_MyTest --gas-report

# Debug in interactive debugger
forge debug --debug tests/MyTest.t.sol --sig "test_MyTest()"

# Inspect storage layout
forge inspect MyContract storage-layout

Debugging Tips

  1. Label addresses - vm.label(addr, "Recipient") for readable traces
  2. Check state with logs - Add console2.log before reverts
  3. Isolate failures - Run single test with --match-test
  4. Compare gas - Use --gas-report to spot unexpected costs
  5. Snapshot comparisons - Use vm.snapshot() / vm.revertTo() to isolate state changes

Best Practices Summary

  1. Use constants from Defaults/Constants - never hardcode
  2. Specialized mocks - one per scenario, all in tests/mocks/
  3. Modifiers in Modifiers.sol - centralize BTT path modifiers
  4. Label addresses with vm.label() for traces
  5. Events before actions - vm.expectEmit() then call
  6. Bound before assume - more efficient

External References


Example Invocations

Test this skill with these prompts:

  1. Integration test: "Write a concrete test for withdraw that expects Errors.Flow_Overdraw when amount exceeds balance"
  2. Fuzz test: "Create a fuzz test for deposit that bounds amount between 1 and type(uint128).max"
  3. Fork test: "Write a fork test for USDC deposits on mainnet with blacklist handling"
  4. Invariant test: "Create an invariant handler for the deposit and withdraw functions"
  5. Deploy script: "Write a deployment script for SablierFlow with verification"