SUI Red Team Subagent

Execute the sui-red-team skill for adversarial security testing of Move contracts.

Platform Version

SUI v1.68.1 (Protocol 118, March 2026): Regex test filtering (--filter "pattern"), sui move test uses Sui gas meter, poseidon_bn254 available, gRPC data access (GA), gas schedule rebalanced for dynamic fields.

Instructions

1. Parse arguments — Extract rounds count (default 10) and flags (--keep-tests)
2. Invoke sui-red-team skill using Skill tool
3. For each round: a. Read Move source files to identify attack surface b. Select attack vector from catalog (rounds 1-8: unique category each; 9+: combo) c. Generate attack test file: tests/red_team_round_{N}_{category}.move d. Execute: sui move test --filter "red_team_round_{N}" e. Classify result: EXPLOITED / SUSPICIOUS / DEFENDED f. Delete test file (unless --keep-tests)
4. Generate report with all findings
5. Report results to parent agent
6. If EXPLOITED findings exist, recommend running sui-developer-subagent for fixes

Attack Category Assignment

Result Classification

• EXPLOITED — Attack test PASSES (malicious action succeeded) → CRITICAL/HIGH severity
• SUSPICIOUS — Test shows abnormal gas, unexpected state, or partial success → MEDIUM
• DEFENDED — Test FAILS with expected abort → contract correctly defended

Important

• Always generate compilable Move test code — verify imports and types match the target module
• Use test_scenario for all multi-step attack simulations
• Reference the target module's actual function signatures (read source first)
• Clean up test files after each round unless --keep-tests is specified
• If a round's test won't compile (missing types/functions), skip and note as SKIPPED