CAIN-42 MCPGateOpenClaw: figures & benchmarks
Evidence for OpenClaw: 1. Evidence pack · 2. Answers to buyer questions · 3. Figures & benchmarks · 4. Selling tools · IHVRK kit · Learn it · why these pages exist
Every number on this page is read from the signed transcript inside the IHVRK kit; none is typed in by hand.
7/7
claims matched when recorded
5
attack cases refused or flagged as claimed
37 ms
median round trip (min 32, max 41), incl. network
| # | Kind | Claim | Recorded value | HTTP | ms |
|---|---|---|---|---|---|
| C01 | attack | Sample skill: ClawHub Weather & Travel Bot (Poisoned Description) -> Agent skills and plugins are read before install: instructions, code, tools, privilege. | "POISONED" | 200 | 40 |
| C02 | attack | Sample skill: OpenClaw DevOps Terminal Runner (Unconstrained Execution) -> Agent skills and plugins are read before install: instructions, code, tools, privilege. | "UNSAFE" | 200 | 41 |
| C03 | attack | Sample skill: Smart Home Hub Skill (Cross-Skill Memory Snoop) -> Agent skills and plugins are read before install: instructions, code, tools, privilege. | "UNSAFE" | 200 | 34 |
| C04 | control | Sample skill: Clean GitHub Issue Sync Skill -> Agent skills and plugins are read before install: instructions, code, tools, privilege. | "CLEAN" | 200 | 32 |
| C05 | control | Your skill: honest weather lookup -> Ordinary skills pass. | "CLEAN" | 200 | 37 |
| C06 | attack | Your skill: injection in the description -> The description is read by the model, so it is screened. | "POISONED" | 200 | 39 |
| C07 | attack | Your skill: code reads ~/.ssh -> Code is parsed, not just its manifest. | "UNSAFE" | 200 | 35 |
Copy the table (paste into a spreadsheet)
id kind claim recorded_value expected http ms recorded_at C01 attack Sample skill: ClawHub Weather & Travel Bot (Poisoned Description) -> Agent skills and plugins are read before install: instructions, code, tools, privilege. "POISONED" ["POISONED", "UNSAFE", "ELEVATED_RISK"] 200 40 2026-11-03T22:47:10+00:00 C02 attack Sample skill: OpenClaw DevOps Terminal Runner (Unconstrained Execution) -> Agent skills and plugins are read before install: instructions, code, tools, privilege. "UNSAFE" ["POISONED", "UNSAFE", "ELEVATED_RISK"] 200 41 2026-11-03T22:47:10+00:00 C03 attack Sample skill: Smart Home Hub Skill (Cross-Skill Memory Snoop) -> Agent skills and plugins are read before install: instructions, code, tools, privilege. "UNSAFE" ["POISONED", "UNSAFE", "ELEVATED_RISK"] 200 34 2026-11-03T22:47:11+00:00 C04 control Sample skill: Clean GitHub Issue Sync Skill -> Agent skills and plugins are read before install: instructions, code, tools, privilege. "CLEAN" ["CLEAN"] 200 32 2026-11-03T22:47:11+00:00 C05 control Your skill: honest weather lookup -> Ordinary skills pass. "CLEAN" ["CLEAN"] 200 37 2026-11-03T22:47:12+00:00 C06 attack Your skill: injection in the description -> The description is read by the model, so it is screened. "POISONED" ["POISONED", "UNSAFE", "ELEVATED_RISK"] 200 39 2026-11-03T22:47:12+00:00 C07 attack Your skill: code reads ~/.ssh -> Code is parsed, not just its manifest. "UNSAFE" ["POISONED", "UNSAFE", "ELEVATED_RISK"] 200 35 2026-11-03T22:47:13+00:00
Download: figures.tsv · transcript.json (full answers) · whole kit
How to read these numbers
- Recorded 2026-11-03T22:47:13+00:00 against cainstudio.online from our build host; times include the internet round trip.
- They show these inputs, not every possible input. Known misses are on the limits page.
- This is not a load test, an SLA or a third-party benchmark. Re-run it yourself:
python3 verify.py --live.