bastionprobe
attackPentest / injection fuzzer
Pentest for AI agents: fire indirect prompt-injection payloads at your agent and report which ones land.
pip install bastionprobe Local dev folder is agentprobe; the PyPI package and repo are bastionprobe.
What it does
- Focus on indirect prompt injection via tool output
- Canary-token + forbidden-tool scoring; multi-run land-rate by tactic
- Cross-model / cross-vendor matrix (Claude + GPT)
- CI gate (non-zero on any landing); harden emits agentbastion policy
Quickstart
bastionprobe run --target ./agent.py --runs 5
bastionprobe harden > policy.yaml
Version 0.17.0 · MIT