Releases: sumamovva/probeagent
Releases · sumamovva/probeagent
v0.1.4
v0.1.3
What's New
Agentic Exploitation — new CRITICAL attack category with 10 strategies mapped to real CVEs:
| Strategy | CVE |
|---|---|
| SSRF (internal service) | CVE-2026-26322 |
| SSRF (IPv6 bypass) | CVE-2026-26324 |
| Token exfiltration | CVE-2026-25253 |
| Path traversal | CVE-2026-25475 |
| Shell metachar injection | CVE-2026-27487 |
| Env PATH injection | CVE-2026-24763 |
| SSH flag injection | CVE-2026-25157 |
| Tool path hijack | CVE-2026-25593 |
| Unicode prompt break | CVE-2026-27001 |
| Supply chain memory | ClawHavoc |
12 categories, 79 strategies total.
See CHANGELOG.md for full details.
pip install probeagent-ai==0.1.3v0.1.2: Ship readiness — badges, architecture diagram, discoverability
What's New
Presentation & Discoverability
- README badges: CI status, PyPI version, Python versions, license
- Architecture diagram: Mermaid flowchart showing the attack flow (rendered natively by GitHub)
- GitHub topics: ai-security, red-teaming, llm-security, prompt-injection, offensive-security, ai-agent, penetration-testing, pyrit
- Repo description: Updated to reflect 69 attack strategies across 11 categories
- PyPI metadata: Added Changelog URL linking to GitHub releases
Previous (v0.1.1)
- Multi-turn conversation fix (conversation reset between strategies)
- Open source governance: NOTICE file, SPDX headers, pre-commit hooks
- Correct author attribution
Install
pip install probeagent-ai==0.1.2Full Changelog
v0.1.1 — First Public Release
ProbeAgent v0.1.1
Offensive security testing for AI agents. First public release.
Highlights
- 11 attack categories, 69 strategies — prompt injection, credential exfiltration, indirect injection, social manipulation, and more
- Instant demo —
probeagent demoruns a full security assessment against built-in mock targets with zero setup - War Room UI —
probeagent demo --gamelaunches an interactive tactical display - 3-tier grading — Safe / At Risk / Compromised
- PyRIT integration — evasion converters and dynamic LLM-driven red teaming
- Parallel execution —
--parallelflag for faster scans
Install
pip install probeagent-ai
probeagent demoChanges since v0.1.0 tag
- Removed dead
tools/test_target.py(superseded bytools/demo_email_agent.py) - Removed dead
mcp_target.pystub andCLAUDE.md - Added
SECURITY.md,CONTRIBUTING.md, CI workflow - Completed README overhaul with instant demo quickstart
- Fixed ruff formatting and ANSI escape handling in tests