AegisGate Lens v0.2.0

AegisGate Lens v0.2.0

Release date: 2026-07-27
Tag: v0.2.0

What's new in v0.2.0

OPSEC hardening

• Pre-commit hook (.githooks/pre-commit): detects private keys, local paths, internal doc references, CWS IDs, hardcoded secrets, binaries, and .env files before they reach the repo.
• CI scan job (opsec-scan in security.yml): automated OPSEC scanning on every push and PR.
• Unified OPSEC scanner (tools/opsec-scan.sh): 12-check scanner matching the CI job.
• Key rotation: CWS screenshot mock TLS key rotated from RSA-2048 to ECDSA P-256.

OPSEC sweep

• Removed all private keys, internal docs, build artifacts, and local paths from the public repo (PR #40).

Bearer token support

• Lens now supports bearer token authentication for self-hosted AegisGate Platform backends (PR #44).

Social links

• Added X/Twitter (@AegisGate) and Mastodon (@aegisgate@mastodon.social) to README and SECURITY docs.

Version bump

• All version strings bumped to v0.2.0 (manifest, source, docs).

Detection metrics (unchanged from v0.1.4)

• 4 detection facets: PII, secrets, XSS, compliance
• Sub-millisecond detection latency
• 100% on-device, zero network egress by default

Privacy & Security

• 12 privacy non-negotiables (no prompt text, no URLs, no page content, no PII, no credentials, no fingerprinting, no cross-site tracking, no AI provider metadata, no keystroke timing, no mouse movement, no session IDs, no IP addresses)
• Apache 2.0, zero external dependencies
• Strict MV3 CSP (no inline scripts, no remote code, no eval)
• All commits GPG-signed
• Vulnerability disclosure: RFC 9116, security@aegisgatesecurity.io

Supported providers

ChatGPT, Claude, Gemini, Microsoft Copilot, DuckDuckGo (Duck.ai), Perplexity, Mistral Le Chat, Grok.

Installation

• Chrome Web Store: search "AegisGate Lens" or visit the listing page
• Manual: unzip, load unpacked in chrome://extensions (Developer mode)

---

SHA-256 and binary assets available in the v0.2.0 tag.

v0.1.4 - Bug #4 fix + 3 v0.1.4 features

AegisGate Lens v0.1.4

Release date: 2026-07-13 (pending CWS approval)
Branch: v0.1.3
HEAD: fdc86a9
Commits: 30 v0.1.4 commits on the v0.1.3 branch (5 session commits, 25 from prior session)

What's new in v0.1.4

Clickable "🛡️ Lens active" indicator (Bug #4 fix)

The "Lens active" indicator chip on every AI chat page is now clickable. Clicking it opens the extension popup, which exposes the 3 v0.1.4 features below. Previously, clicking the chip only logged a console message and did nothing useful.

3 new popup controls (G1, G2, G5)

• G1: "Hide Lens active indicator" — toggle to hide the 🛡️ chip entirely
• G2: "Pause Lens for 1h / 1d" — temporarily disable detection for 1 hour or 1 day
• G5: "Not PII — dismiss for 24h" — dismiss a false-positive detection for 24 hours

Quality infrastructure

• Lightweight shell linter (tools/lint.sh, 12 checks): catches console.log, var, ==, TODO, eval, dynamic innerHTML, missing JSDoc, trailing whitespace, line length > 120, tabs, double-blank-lines, and debugger statements. Runs in ~5s. No npm required.
• Mini-smoke binary (128 tests, 8 hosts × 16 cases): exercises the full Lens flow on each of the 8 AI providers. 5/5 stable, gate=true, ec=0.
• F-7 e2e test category (20 tests): MV3 manifest validation, CSP strictness, permission bounds, 3-way host consistency, bundle freshness checks.
• Performance baseline benchmark: p50=0.08ms (500 chars), p99=0.36ms (2000 chars), p99=1.37ms (10k chars).
• 3-job split in smoke.yml CI: smoke + mini-smoke + lint, all running in parallel.

Test counts (in sync with FACTS.md)

• Node: 500/500 passing (was 477, +20 from F-7 e2e tests, +3 from Bug #4 bundle-freshness check)
• Go: 3/3 passing
• v0.1.3 headless smoke: 5/5 stable, 16/16 each, gate=true
• mini smoke (per-host, 128 tests): 5/5 stable, 128/128 each, gate=true
• Linter: 12/12 checks PASS, 0 warnings, 0 failures

Detection metrics

• 132 regex patterns across 4 detection facets (PII=55, secrets=41, XSS=12, compliance=24)
• 2.31% false positive rate on 6,500 WildChat prompts (5.1× better than v0.1.0-beta baseline)
• 0/119 false positives on the per-pattern must-not-trigger corpus
• Sub-millisecond detection (avg 0.34ms; p99 0.36ms for 2000 chars, p99 1.37ms for 10k chars)
• 5,348 prompts/sec throughput

Privacy & Security

• 12 privacy non-negotiables (no prompt text, no URLs, no page content, no PII, no credentials, no fingerprinting, no cross-site tracking, no AI provider metadata, no keystroke timing, no mouse movement, no session IDs, no IP addresses)
• 100% on-device, zero network egress by default
• Apache 2.0, zero external dependencies (no npm, no node_modules, no bundled libraries)
• Strict MV3 CSP (no inline scripts, no remote code, no eval)
• All commits GPG-signed
• Vulnerability disclosure: RFC 9116 compliant, security@aegisgatesecurity.io

8 AI providers supported

ChatGPT, Claude, Gemini, Microsoft Copilot, DuckDuckGo (Duck.ai), Perplexity, Mistral Le Chat, Grok.

Bug fixes

• Bug #4: "Lens active" indicator is now clickable, opens extension popup
• Bug #4 followup: the original Bug #4 fix's SW handler was outside the IIFE and would have overwritten the main SW onMessage handler in production; now properly integrated into the main switch

What's NOT in v0.1.4 (deferred to v0.2.0)

• Firefox / Edge support (Phase 1, 3-4 weeks)
• Lighthouse CI integration
• Public benchmark dataset release
• Third-party security audit (Cure53 / Trail of Bits / NCC Group)
• Marketing site refresh (aegisgatesecurity.io/lens/)
• ML-based detection facets (toxicity, prompt-injection) — burned down 2026-06-28

Installation

# Chrome Web Store: search for "AegisGate Lens" or visit the developer page
# Manual install (for beta testers):
# 1. Download lens-0.1.4-lens-sr.zip from the Assets section below
# 2. Unzip to a local directory
# 3. Chrome -> chrome://extensions -> enable Developer mode
# 4. Click "Load unpacked" -> select the unzipped directory

SHA-256

lens-0.1.4-lens-sr.zip: a67a83676f39e20e1959067dc930a2dccea59847afbaf1c7da6deffb2d2f2589 (1,326,193 bytes, 59 files)

License

Apache 2.0. Copyright 2026 AegisGate Security, LLC.

Reporting issues

• Bug reports: aegisgate-lens/.github/ISSUE_TEMPLATE/bug_report.md
• Feature requests: aegisgate-lens/.github/ISSUE_TEMPLATE/feature_request.md
• Security disclosures: security@aegisgatesecurity.io (GPG key on request)