Due Diligence Agents

v1.10.0 Security

This release includes 2 security fixes for security teams reviewing exposed deployments.

Published 1mo AI Agents & Assistants

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

This release patches 2 known CVEs

Topics

ai-agents anthropic claude compliance contract-analysis corp-dev

+14 more

corporate-development cross-domain-analysis security document-analysis due-diligence fintech knowledge-graph legal-tech llm mergers-and-acquisitions multi-agent private-equity python risk-analysis

Affected surfaces

auth deps

ReleasePort's take

Moderate signal

editorial:auto 1mo

The trust-layer now includes deterministic tamper detection and a quote‑fidelity gate for hardened security.

Why it matters: Hardens the trust‑layer with deterministic tamper detection; severity score 90 signals critical importance for SREs and security engineers monitoring integrity guarantees.

Summary

AI summary

Updates Changes since v1.9.0, evals, and trust across a mixed release.

Changes in this release

Type	Severity	Summary	CVE
Security	Critical	Hardens trust-layer with deterministic tamper detection and quote-fidelity gate. Hardens trust-layer with deterministic tamper detection and quote-fidelity gate. Source: llm_adapter@2026-06-05 Confidence: high	—
Feature	Medium	HR specialist surfaces each workforce risk as a distinct finding. HR specialist surfaces each workforce risk as a distinct finding. Source: llm_adapter@2026-06-05 Confidence: high	—
Dependency	Low	Update Homebrew formula to version 1.9.0. Update Homebrew formula to version 1.9.0. Source: llm_adapter@2026-06-05 Confidence: high	—
Performance	Low	Increase eval F1 no‑regression tolerance from 0.05 to 0.15. Increase eval F1 no‑regression tolerance from 0.05 to 0.15. Source: llm_adapter@2026-06-05 Confidence: high	—
Bugfix	Medium	Treat agent timeout/non-success as INCONCLUSIVE instead of 0.0 recall. Treat agent timeout/non-success as INCONCLUSIVE instead of 0.0 recall. Source: llm_adapter@2026-06-05 Confidence: high	—
Bugfix	Medium	Close three masking holes discovered by competitive audit of fa66531. Close three masking holes discovered by competitive audit of fa66531. Source: llm_adapter@2026-06-05 Confidence: high	—

Full changelog

Changes since v1.9.0

v1.10.0: trust-layer hardening + eval-robustness overhaul
test(evals): size F1 no-regression tolerance to real variance (0.05 -> 0.15)
test(evals): authoritative median-of-3 baseline — all 9 specialists pass live
test(evals): treat agent timeout/non-success as INCONCLUSIVE, not 0.0 recall
feat(agents): HR specialist surfaces each workforce risk as a distinct finding
test(evals): capacity-K (K=2) recall matching — credit consolidation, cap stuffing
test(evals): close 3 masking holes found by competitive audit of fa66531
test(evals): make agent evals reliable and cover all 9 specialists (no masking)
test(trust): cover engine wiring for tamper injection + numerical-audit text_dir
docs(devto): update trust-layer article for shipped deterministic gates
feat(trust): wire deterministic tamper detection + add quote-fidelity gate
content: add dev.to article on the trust layer (audited + verified)
brand: add iris logo — README header + MkDocs theme logo/favicon
docs: add "How the Agents Work" — an audit-first agent anatomy tour
Update Homebrew formula to 1.9.0

Install

pip install dd-agents==1.10.0

Docker

docker pull zoharbabin/due-diligence-agents:1.10.0

Full Changelog: https://github.com/zoharbabin/due-diligence-agents/compare/v1.9.0...v1.10.0

Security Fixes

test(trust): cover engine wiring for tamper injection + numerical-audit text_dir
test(evals): close 3 masking holes found by competitive audit of fa66531

View diff on GitHub

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

Share on X Share on Bluesky

Track Due Diligence Agents

Get notified when new releases ship.

About Due Diligence Agents

All releases →