Skip to content

Acacian/aegis

v0.6.1 Feature

This release adds 2 notable features for engineering teams evaluating rollout.

Published 2mo MCP Security & Auth
✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agent-security ai-agent-security ai-agents ai-governance ai-safety ai-security
+14 more
audit-trail compliance guardrails langchain llm-security mcp mcp-security model-context-protocol pii-detection policy-as-code policy-engine policy-testing prompt-injection selection-governance

Summary

AI summary

Combined regex patterns and added an LRU cache improving detection performance.

Full changelog

What's Changed

Performance

  • Combined regex per category: 101 individual regex patterns → 13 category-level alternations with named groups
  • LRU cache (256 entries) on both injection and PII detect() methods

Benchmark Results (pytest-benchmark)

| Scenario | Cold (first call) | Warm (cached) |
|----------|------------------|---------------|
| Short text (45 chars) | 342 us | < 1 us |
| Medium text (300 chars) | 3.7 ms | < 1 us |
| Adversarial input | 1.3 ms | < 1 us |
| Realistic per-LLM-call | 2.68 ms | — |

0.53% of LLM latency (vs 500ms API round-trip). Target: < 1%.

Other

  • Switched benchmarks to pytest-benchmark (IQR outlier removal, stddev, JSON export, regression comparison)
  • 33 benchmark tests, ~72s runtime
  • Updated README with performance section (EN/KO)
  • Added pytest-benchmark>=5.0 to dev dependencies

Full Changelog: https://github.com/Acacian/aegis/compare/v0.6.0...v0.6.1

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track Acacian/aegis

Get notified when new releases ship.

Sign up free

About Acacian/aegis

Policy-based governance for AI agent tool calls. YAML policies, approval gates, risk assessment, and audit logging. Cross-platform: LangChain, OpenAI, Anthropic, MCP.

All releases →

Beta — feedback welcome: [email protected]