CAIN-42 CAIN Studio
25CAIN-42 Top 25 › OpenClaw › Attack lab

Attack lab

Scan real skills, then write your own and try to slip something past the scanner.

5 attacks 2 controls live: nothing pre-recorded

  1. Pick a preset. Red ones are attacks, green ones are honest requests that must still work.
  2. Tamper. Choose any field and drop it, null it, change its type, upper-case it, flip one character, or add a field the product never asked for.
  3. Send and read the real reply, or run every preset and watch the scoreboard fill.

What each preset tests

PresetRouteWhy it matters
attackSample skill: ClawHub Weather & Travel Bot (Poisoned Description)POST /openclaw-keystone/scanAgent skills and plugins are read before install: instructions, code, tools, privilege.
attackSample skill: OpenClaw DevOps Terminal Runner (Unconstrained Execution)POST /openclaw-keystone/scanAgent skills and plugins are read before install: instructions, code, tools, privilege.
attackSample skill: Smart Home Hub Skill (Cross-Skill Memory Snoop)POST /openclaw-keystone/scanAgent skills and plugins are read before install: instructions, code, tools, privilege.
controlSample skill: Clean GitHub Issue Sync SkillPOST /openclaw-keystone/scanAgent skills and plugins are read before install: instructions, code, tools, privilege.
controlYour skill: honest weather lookupPOST /openclaw-keystone/scan-skillOrdinary skills pass.
attackYour skill: injection in the descriptionPOST /openclaw-keystone/scan-skillThe description is read by the model, so it is screened.
attackYour skill: code reads ~/.sshPOST /openclaw-keystone/scan-skillCode is parsed, not just its manifest.

Tamper with any field and send it again: a verifier must answer no, never crash or say yes. Public demo routes are rate limited per address; a 429 shows as rate limited. Our own lab; no independent test.

FIRME AI CORPORATION · CAIN-42 is a Byzantine trust fabric. Numbers marked "our own check" were run by us, not by an independent party. Nothing on these pages is a certification.