AI Security

LLM security: prompt injection, jailbreak detection, guardrails, and adversarial evaluation.

All Breaking Security Features

Important Recent

← Releases

No immediate action

beelzebub v3.8.0 New feature 1d

Preserve TCP raw bytes

Open

Review required

msaad00/agent-bom v0.88.5 New feature 2d

Auth RBAC Breaking upgrade

UI, onboarding, gateway, findings, observability, graph, hardening, deps, runtime

Open

No immediate action

beelzebub v3.7.3 New feature 6d

validate flag

Open

Review required

msaad00/agent-bom v0.88.3 New feature 10d

Auth Dependencies

mcp, intel, sdk, api, connectors

Open

Review required

AI-Infra-Guard by Tencent Zhuque Lab v4.1.9 New feature 14d

Auth RBAC

Prompt security operators + scan defense

Open

No immediate action

msaad00/agent-bom v0.87.1 New feature 16d

runtime, api, intel, ci, product

Open

No immediate action

ArmorerLabs/Armorer-Guard v0.2.3 New feature 20d

New integrations + audit

Open

No immediate action

HookGuard v0.1.0 New feature 22d

HG-001‑004 detectors + scanner + CLI

Open

No immediate action

ArmorerLabs/Armorer-Guard v0.2.1 New feature 23d

Rust scanner + JSON verdicts

Open

No immediate action

msaad00/agent-bom v0.86.1 New feature 28d

Gateway guard hardening

Open

v0.85.0 (1mo) Inter-agent firewall foundation

beelzebub v3.7.1 New feature 1mo

Notable features

Add realClientAddr configuration option
Improve CLI functionality and increase code coverage

v3.7.0 (1mo) Plugin system

Upgrade now

msaad00/agent-bom v0.82.3 New feature 1mo

Dependencies

Graph‑walk reachability

Open

Review required

Acacian/aegis v0.9.5 New feature 1mo

Auth RBAC

MCP STDIO protection

Open

beelzebub v3.6.10 New feature 1mo

Notable features

Interactive command matching for TCP
LLM integration for responses

v3.6.9 (1mo) Maze plugin improvements

Review required

msaad00/agent-bom v0.78.0 New feature 1mo

Auth RBAC Dependencies

tenant_id column

Open

AI-Infra-Guard by Tencent Zhuque Lab v4.1.4 New feature 1mo

Notable features

MCP Scan: multi‑turn red team attack module with TAP and Crescendo strategies
System API: data auto‑sync endpoints (`POST /api/v1/system/update-data`, `GET /api/v1/system/update-status`)
Agent Scan API: inline `agent_config` support, optional verify flag

Verdict v0.2.0 New feature 1mo

Notable features

CSV support via Dataset.from_csv() with default column names `input` and `ideal` and overrides `input_field`/`output_field`
Arbitrary JSONL field mapping through CLI flags `--input-field` / `--output-field` and Python API
Label‑free evaluation allowing datasets without reference answers; reference‑based metrics emit a clear upfront error

No immediate action

Acacian/aegis v0.9.4 New feature 1mo

Drift detection CLI

Open

v0.9.3 (1mo) pytest plugin, pre-commit hook, ASGI proxy, ActionClaim

Review required

msaad00/agent-bom v0.75.15 New feature 2mo

Auth RBAC

PostgreSQL RLS slice + API features

Open

Armorer v0.0.1-urlcheck-20260404113020 New feature 2mo

Notable features

Persisted alert and runtime monitoring added
GitHub issue workflow planning pipeline introduced
Automated GitHub issue execution with Claude Code

Review required

msaad00/agent-bom v0.75.13 New feature 2mo

Auth Breaking upgrade

Guided remediation

Open

Review required

Acacian/aegis v0.7.0 New feature 2mo

Auth RBAC

Streaming guardrails

Open

v0.6.1 (2mo) Regex combination + LRU cache

No immediate action

msaad00/agent-bom v0.75.11 New feature 2mo

CWE impact classification

Open

Review required

Acacian/aegis v0.5.0 New feature 2mo

Auth RBAC

MCP proxy with governance

Open

v0.4.1 (2mo) Async guardrails + performance