菜单
How to run the Swoosh test suite efficiently. Use when running swift test, narrowing to a single target, debugging a "hang at 100% passed" runner, or asked "did I break X". Encodes the multi-filter hang limitation, the cooperative-pool deadlock pattern fixed in e087609, and expected walltimes.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Multi-file checklist for adding a new SwooshPermission case. Use when adding a permission, exposing a new capability that needs firewall gating, or wiring a new Scout source / tool family / API surface that requires its own grant. Encodes the enum + profile grants + docs + tool wiring + test path so nothing is forgotten.
Interact with Jupiter Lend Protocol. Read-only SDK (@jup-ag/lend-read) for querying liquidity pools, lending markets (jlTokens), and vaults. Write SDK (@jup-ag/lend) for lending (deposit/withdraw) and vault operations (deposit collateral, borrow, repay, manage positions).
Migration guide from Jupiter Metis (v1) or Ultra to Swap API v2. Use when migrating existing Jupiter swap integrations, updating base URLs, or transitioning from quote+swap-instructions to the unified build endpoint.
Use when a user mentions Jupiter token verification, VRFD eligibility, paying 1000 JUP to verify a token, submitting a verification request, or updating metadata via the Jupiter express verification flow.
Use Bags for Solana token launch planning, agent authentication, launch intent creation, and launch transaction creation through the official Bags API.
Use Flap for BNB Chain token trading, wallet and bot integrations, token-launcher flows, VaultPortal launches, deployed contract references, and Blink-style surfaces.
| name | run-swoosh-tests |
| description | How to run the Swoosh test suite efficiently. Use when running swift test, narrowing to a single target, debugging a "hang at 100% passed" runner, or asked "did I break X". Encodes the multi-filter hang limitation, the cooperative-pool deadlock pattern fixed in e087609, and expected walltimes. |
~50 test targets, 1749 swift-testing tests, ~10–15s walltime after a fresh build (~2s incremental). Full swift test works cleanly as of e087609.
| Goal | Command |
|---|---|
| "Does it compile" | swift build |
| "Is the tree healthy" | swift test |
| "Did I break module X" | swift test --filter <Module>Tests |
| One specific test | swift test --filter "<Target>.<Suite>/<test>" |
| Discover test IDs | swift test --list-tests | grep <substr> |
--filter takes the test ID from --list-tests, not the display name. "Initialize completes" will match nothing.
--filter flags in one invocation hang at startup with 0% CPU. Loop in shell instead:
for t in SwooshFirewallTests SwooshAgentLoopTests SwooshFlowTests; do
echo "=== $t ==="
swift test --filter $t 2>&1 | grep "Test run" | tail -1
done
Symptom: tests all print "passed", then the runner sits at 0% CPU indefinitely and never exits.
Recipe:
ps -ef | grep swift-test to get the PID.sample <PID> 2 -file /tmp/sample.txt && head -200 /tmp/sample.txthandle.read(upToCount:) or any blocking syscall. Any blocking read inside Task.detached will starve the pool — fix by moving it to FileHandle.readabilityHandler (callback runs on FileHandle's private dispatch queue). See Sources/SwooshMCP/MCPTransport.swift for the canonical fix.NIO-SGLTN-1-#N threads parked in kevent — those are SwiftNIO's process-wide singleton EventLoopGroup, not a hang cause.swift test --parallel — the default already parallelizes; the flag has historically destabilized the stdio MCP path.xcodebuild test for library/CLI/daemon work — it's slower and won't pick up package-level changes without scheme regeneration. Use swift test.swift test --enable-code-coverage slows the run significantly; reserve for explicit coverage passes.