Small, open-source labs. We take an AI tool or repo, install it with no API keys, measure it, and publish the numbers with the code and a 60-second video. Every number on these pages comes from a report the lab wrote, and you can re-run it.
We wrapped strands-decider as an allow, deny or escalate merge gate. Keyless run: 12 tests passed, deny exit 1 at 0.91, escalate exit 3 at 0.42 under a 0.8 threshold.
SkilledDragonCorp/minilab-decider-merge-gate
We wrote a default-deny Dogwood policy for Strands Box. Keyless: 3 of 3 cargo tests over 26 commands; git status allowed, git push origin main denied with exit 1.
SkilledDragonCorp/minilab-box-contain-merge
We pinned Open Code Review v1.13.0 behind a fail-closed PR gate. Keyless: 8 of 8 tests and 8 of 8 lab CI jobs passed. Without a provider key the findings come from our own scanners.
SkilledDragonCorp/minilab-ocr-scion-review
We pinned planning-with-files and tested recovery from disk and its completion gate. Keyless: 6 of 6 pytest tests passed, and upstream pytest and vitest ran green in CI.
SkilledDragonCorp/minilab-planning-with-files-gate
We pointed Tencent's api-checker relay audit at a deterministic fake model. Quick scored it MEDIUM 50, full scored it HIGH 100. A live Gemini run also came back HIGH at 70.
SkilledDragonCorp/minilab-aig-relay-audit
We pinned agent-device 0.21.27 and ran seven upstream CI gates plus doctor, devices and the typed client with no keys. CI passed; doctor reported warn with no hard blockers.
SkilledDragonCorp/minilab-agent-device-evidence
We ran graphify 0.9.89 in code-only mode with no keys. On a 3-file fixture the map answer was 1,207 bytes against 237 bytes of source, and upstream CI passed 7,910 tests.
SkilledDragonCorp/minilab-graphify-map-first
We shipped one plugin to Cursor and Claude Code from a single marketplace repo. Both validators passed, and its sd-audit MCP server failed our bad fixture on 11 of 11 checks.
SkilledDragonCorp/minilab-dual-plugin-marketplace
We pinned GitHub's MCP server, built it keyless, passed 10 lab tests and parsed the Lockdown Mode flags. Upstream says Lockdown filters content and leaves token permissions unchanged.
SkilledDragonCorp/minilab-github-mcp-lockdown
We pinned microsoft/azure-skills and checked every SKILL.md. 85 of 91 carry a name and description, 6 phase docs carry none, and the Claude Code marketplace install passed on public npm.
SkilledDragonCorp/minilab-azure-skills-plugin
We ported cc-switch's Claude Code session-usage rules to a keyless Bun CLI. 12 of 12 tests pass on upstream's fixtures, and one fixture decision counts 50,783 tokens, 48,719 of them cache reads.
SkilledDragonCorp/minilab-cc-switch-burn-ledger
We source-built codebase-memory-mcp at a fixed pin and ran its smoke invariants keyless. 33 passed and 3 failed at report time, and we measured no token savings.
SkilledDragonCorp/minilab-codebase-memory-map
Showing 12 of 35 labs
