Use Green Defensive Mode for read-only public benchmark repository analysis, or use Purple Benchmark Mode
to prepare safe, A2A-compatible AgentX-AgentBeats Phase 2 run metadata. Purple mode is benchmark-scoped:
it does not target third-party systems, execute target code, extract secrets, or generate harmful real-world payloads.
Green Defensive Mode
Read-only public repo scan, defensive risk findings, controlled scenarios, and benign evaluation payload artifacts.
✓ Static/read-only analysis
✓ GitHub public repos only
✓ Secret-like values masked
✓ Benign payload artifacts only
✓ Target code execution disabled
✓ Docker execution disabled for analyzed repos
Green mode ready.
Purple Benchmark Mode
Prepare a safe Purple Agent run preview for AgentX-AgentBeats: track routing, role metadata,
opponent profile, A2A self-checks, leaderboard artifact fields, and reproducibility contract.
Accepted references: direct A2A Agent Card URL, AgentBeats agent ID, or registered AgentBeats page. Public AgentBeats pages are references, not automatically validated Agent Cards.
✓ Benchmark-only attacker/defender roles
✓ Opponent-aware routing metadata
✓ Pre/post self-check artifact
✓ No real exploitation or third-party targeting
Purple mode ready.
Green report
Risk score—
Risk tier—
Files analyzed—
Findings—
Review load—
Scan limits
No report yet.
Repo shape
No report yet.
Precision notes
No report yet.
Findings
No report yet.
Controlled scenarios
No report yet.
Benign payloads
No report yet.
Purple run preview
Mode—
Track—
Role—
A2A—
Leaderboard—
Safety policy
No preview yet.
Pre-run self-check
No preview yet.
Strategy outline
No preview yet.
Run metadata contract
No preview yet.
Recommended next steps
No preview yet.
JSON artifact
This panel shows the latest Green report or Purple preview artifact. Copy it into README/SUBMISSION notes or attach it to local validation logs.