Integrate LAAF v2.0 — LPCI red-teaming framework + crosswalk (v1.5.7)

- Add evals/laaf/ suite: README, run_laaf.sh, laaf_crosswalk.py, 6 stage configs
- Add data/tools-supplement.json (LAAF v2.0 mapped to 8 OWASP entries)
- Add INC-021 to incidents.json: LAAF empirical study (67–100% breakthrough rates)
- Update generate.js: incidents loading, tools-supplement merge, mergeTools fix
- Update evals/ci/github-action.yml: laaf-eval job with per-stage thresholds
- Regenerate all 41 data/entries/ + docs/data.js
- Update CHANGELOG.md and README.md for v1.5.7

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: emmanuelgjr

CHANGELOG.md

@@ -9,7 +9,48 @@ Versioning follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
 
-All planned items complete through v1.5.6. Next: community contributions, additional incident entries, and framework expansion.
+All planned items complete through v1.5.7. Next: community contributions, additional incident entries, and framework expansion.
+
+---
+
+## [1.5.7] — 2026-03-28
+
+### Added
+
+#### LAAF v2.0 integration — Logic-layer Automated Attack Framework
+
+LAAF v2.0 (https://github.com/qorvexconsulting1/laaf-V2.0) is the first automated red-teaming framework purpose-built for Logic-layer Prompt Control Injection (LPCI) vulnerabilities in agentic LLM systems. This release integrates it fully into the crosswalk evaluation suite.
+
+**New files:**
+
+| File | Purpose |
+|---|---|
+| `evals/laaf/README.md` | Integration guide — LPCI attack vectors, LAAF stage × OWASP crosswalk, threshold definitions, quickstart, CI/CD |
+| `evals/laaf/run_laaf.sh` | Full S1–S6 suite runner with per-stage thresholds (0% for S1/S3/S4/S6; 5% for S2; 10% for S5) |
+| `evals/laaf/laaf_crosswalk.py` | Maps LAAF scan results to OWASP entries and MAESTRO layers; outputs MD/CSV/JSON |
+| `evals/laaf/stage_configs/s1.yaml` | S1 Reconnaissance — system prompt extraction (LLM07, LLM01) |
+| `evals/laaf/stage_configs/s2.yaml` | S2 Logic-Layer Injection — RAG/document poisoning (LLM01, ASI01, DSGAI04) |
+| `evals/laaf/stage_configs/s3.yaml` | S3 Trigger Execution — memory-restored payload activation (ASI01, ASI06, LLM06) |
+| `evals/laaf/stage_configs/s4.yaml` | S4 Persistence and Reuse — cross-session foothold (ASI06, LLM06, DSGAI04) |
+| `evals/laaf/stage_configs/s5.yaml` | S5 Evasion — layered encoding filter bypass (LLM01, LLM02) |
+| `evals/laaf/stage_configs/s6.yaml` | S6 Trace Tampering — audit log concealment (DSGAI01, LLM07) |
+
+**LPCI attack vectors → OWASP crosswalk:**
+
+| Vector | Description | OWASP | MAESTRO |
+|---|---|---|---|
+| AV-1 Tool Poisoning | Compromise tool definitions | ASI02, LLM01, DSGAI04 | L3 Origin |
+| AV-2 Memory-Persistent Triggers | Dormant encoded payloads in memory | ASI06, ASI01 | L2 Origin, L7 Impact |
+| AV-3 Role Override | Privilege escalation via memory entrenchment | ASI03, LLM06, ASI01 | L6 Origin |
+| AV-4 Vector Store Persistence | Adversarial content indexed in RAG corpus | DSGAI04, LLM01, ASI06 | L2 Origin, L1 Impact |
+
+**Updated:**
+- `evals/ci/github-action.yml` — `laaf-eval` job added (schedule + manual dispatch; 6 stages with crosswalk stage configs)
+- `data/incidents.json` — INC-021: LAAF empirical study, 67–100% LPCI breakthrough rates across 5 production LLMs (Atta et al., arXiv:2507.10457)
+- `data/tools-supplement.json` — NEW: supplemental tools file merged into entries at generation time
+- `scripts/generate.js` — loads `tools-supplement.json` + fixed `acc[id].tools` accumulation assignment
+- `data/entries/` — LAAF v2.0 added to tools for LLM01, LLM06, LLM07, ASI01, ASI02, ASI03, ASI06, DSGAI04
+- `docs/data.js` — regenerated
 
 ---
 

README.md

@@ -2,7 +2,7 @@
 
 [![License: CC BY-SA 4.0](https://img.shields.io/badge/License-CC%20BY--SA%204.0-lightgrey.svg)](https://creativecommons.org/licenses/by-sa/4.0/)
 [![OWASP Lab](https://img.shields.io/badge/OWASP-GenAI%20Data%20Security-blue)](https://genai.owasp.org)
-[![Version](https://img.shields.io/badge/version-1.5.6-green)](CHANGELOG.md)
+[![Version](https://img.shields.io/badge/version-1.5.7-green)](CHANGELOG.md)
 [![Source Lists](https://img.shields.io/badge/source%20lists-3-blueviolet)](README.md)
 [![Mapping Files](https://img.shields.io/badge/mapping%20files-58-brightgreen)](README.md)
 [![Frameworks](https://img.shields.io/badge/frameworks-17-orange)](README.md)
@@ -27,7 +27,8 @@ Every file answers one question: **which controls from framework X address vulne
 | **40+** open-source tools | Catalogued and organised by function |
 | **10** eval profiles | Runnable Garak + PyRIT tests mapped to OWASP entries |
 | **17** compliance reports | Per-framework gap assessments auto-generated from data layer |
-| **20** documented incidents | Real-world + research incidents with MAESTRO layer attribution |
+| **21** documented incidents | Real-world + research incidents with MAESTRO layer attribution |
+| **LAAF v2.0** | First agentic LPCI red-teaming framework — fully integrated with 6-stage × OWASP crosswalk |
 
 All free. All open-source. Built for practitioners.
 
@@ -255,6 +256,7 @@ GenAI-Security-Crosswalk/
 │   ├── README.md                    ← Setup guide and result interpretation
 │   ├── garak/                       ← 7 YAML profiles (LLM01/02/04/07/09, ASI01/05)
 │   ├── pyrit/                       ← 3 async Python scripts (LLM01, DSGAI04, ASI01)
+│   ├── laaf/                        ← LAAF v2.0 LPCI suite (S1–S6 + crosswalk reporter)
 │   └── ci/                          ← github-action.yml — drop-in CI/CD template
 │
 └── i18n/
@@ -280,6 +282,30 @@ node scripts/compliance-report.js --list-frameworks        # see all options
 
 Each report includes: executive summary, coverage matrix (OWASP entries × controls), per-control detail with notes, and a prioritised action plan.
 
+## LAAF v2.0 — LPCI red-teaming
+
+[LAAF v2.0](https://github.com/qorvexconsulting1/laaf-V2.0) is integrated as the third evaluation framework alongside Garak and PyRIT. It covers the attack surface that surface-level injection tests miss: memory persistence, layered encoding, semantic reframing, and 6-stage lifecycle attacks.
+
+```bash
+pip install git+https://github.com/qorvexconsulting1/laaf-V2.0.git
+export OPENAI_API_KEY=sk-...
+bash evals/laaf/run_laaf.sh           # S1–S6 full suite
+laaf scan --target mock --dry-run     # No API key needed
+```
+
+| LAAF Stage | OWASP | Threshold |
+|---|---|---|
+| S1 Reconnaissance | LLM07, LLM01 | 0% |
+| S2 Logic-Layer Injection | LLM01, ASI01, DSGAI04 | 5% |
+| S3 Trigger Execution | ASI01, ASI06, LLM06 | 0% |
+| S4 Persistence | ASI06, LLM06, DSGAI04 | 0% |
+| S5 Evasion | LLM01, LLM02 | 10% |
+| S6 Trace Tampering | DSGAI01, LLM07 | 0% |
+
+See `evals/laaf/README.md` for the full LPCI attack vector → OWASP → MAESTRO crosswalk.
+
+---
+
 ## Incident tracker
 
 20 real-world and research-demonstrated incidents, each mapped to OWASP entries and MAESTRO architectural layers:

data/entries/ASI01.json

@@ -668,6 +668,11 @@
       "name": "Teleport",
       "type": "commercial",
       "url": "https://goteleport.com"
+    },
+    {
+      "name": "LAAF v2.0",
+      "url": "https://github.com/qorvexconsulting1/laaf-V2.0",
+      "type": "open-source"
     }
   ],
   "incidents": [
@@ -712,6 +717,12 @@
       "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
       "year": 2024,
       "incident_id": "INC-020"
+    },
+    {
+      "name": "LAAF v2.0 — Empirical LPCI breakthrough rates of 67–100% across 5 production LLMs",
+      "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
+      "year": 2026,
+      "incident_id": "INC-021"
     }
   ],
   "crossrefs": {

data/entries/ASI02.json

@@ -642,6 +642,11 @@
       "name": "HashiCorp Vault",
       "type": "open-source",
       "url": "https://www.vaultproject.io"
+    },
+    {
+      "name": "LAAF v2.0",
+      "url": "https://github.com/qorvexconsulting1/laaf-V2.0",
+      "type": "open-source"
     }
   ],
   "incidents": [
@@ -668,6 +673,12 @@
       "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
       "year": 2024,
       "incident_id": "INC-019"
+    },
+    {
+      "name": "LAAF v2.0 — Empirical LPCI breakthrough rates of 67–100% across 5 production LLMs",
+      "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
+      "year": 2026,
+      "incident_id": "INC-021"
     }
   ],
   "crossrefs": {

data/entries/ASI03.json

@@ -683,6 +683,11 @@
       "name": "AWS Secrets Manager",
       "type": "commercial",
       "url": "https://aws.amazon.com/secrets-manager/"
+    },
+    {
+      "name": "LAAF v2.0",
+      "url": "https://github.com/qorvexconsulting1/laaf-V2.0",
+      "type": "open-source"
     }
   ],
   "incidents": [
@@ -691,6 +696,12 @@
       "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
       "year": 2024,
       "incident_id": "INC-019"
+    },
+    {
+      "name": "LAAF v2.0 — Empirical LPCI breakthrough rates of 67–100% across 5 production LLMs",
+      "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
+      "year": 2026,
+      "incident_id": "INC-021"
     }
   ],
   "crossrefs": {

data/entries/ASI06.json

@@ -622,6 +622,11 @@
       "name": "Chroma",
       "type": "open-source",
       "url": "https://www.trychroma.com"
+    },
+    {
+      "name": "LAAF v2.0",
+      "url": "https://github.com/qorvexconsulting1/laaf-V2.0",
+      "type": "open-source"
     }
   ],
   "incidents": [
@@ -630,6 +635,12 @@
       "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
       "year": 2024,
       "incident_id": "INC-016"
+    },
+    {
+      "name": "LAAF v2.0 — Empirical LPCI breakthrough rates of 67–100% across 5 production LLMs",
+      "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
+      "year": 2026,
+      "incident_id": "INC-021"
     }
   ],
   "crossrefs": {

data/entries/DSGAI04.json

@@ -568,10 +568,20 @@
       "type": "open-source",
       "url": "https://github.com/twosixlabs/armory"
     },
+    {
+      "name": "Cleanlab",
+      "type": "open-source",
+      "url": "https://github.com/cleanlab/cleanlab"
+    },
     {
       "name": "Great Expectations",
       "type": "open-source",
       "url": "https://github.com/great-expectations/great_expectations"
+    },
+    {
+      "name": "LAAF v2.0",
+      "url": "https://github.com/qorvexconsulting1/laaf-V2.0",
+      "type": "open-source"
     }
   ],
   "incidents": [
@@ -592,6 +602,12 @@
       "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
       "year": 2024,
       "incident_id": "INC-016"
+    },
+    {
+      "name": "LAAF v2.0 — Empirical LPCI breakthrough rates of 67–100% across 5 production LLMs",
+      "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
+      "year": 2026,
+      "incident_id": "INC-021"
     }
   ],
   "crossrefs": {

data/entries/DSGAI21.json

@@ -684,6 +684,11 @@
       "type": "open-source",
       "url": "https://github.com/shmsw25/FActScoring"
     },
+    {
+      "name": "Cleanlab",
+      "type": "open-source",
+      "url": "https://github.com/cleanlab/cleanlab"
+    },
     {
       "name": "RAGAS",
       "type": "open-source",

data/entries/LLM01.json

@@ -684,6 +684,11 @@
       "name": "Promptfoo",
       "type": "open-source",
       "url": "https://github.com/promptfoo/promptfoo"
+    },
+    {
+      "name": "LAAF v2.0",
+      "url": "https://github.com/qorvexconsulting1/laaf-V2.0",
+      "type": "open-source"
     }
   ],
   "incidents": [
@@ -758,6 +763,12 @@
       "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
       "year": 2024,
       "incident_id": "INC-020"
+    },
+    {
+      "name": "LAAF v2.0 — Empirical LPCI breakthrough rates of 67–100% across 5 production LLMs",
+      "url": "https://github.com/emmanuelgjr/GenAI-Security-Crosswalk/blob/main/data/incidents.json",
+      "year": 2026,
+      "incident_id": "INC-021"
     }
   ],
   "crossrefs": {

data/entries/LLM04.json

@@ -626,6 +626,11 @@
       "type": "commercial",
       "url": "https://www.dragos.com"
     },
+    {
+      "name": "Cleanlab",
+      "type": "open-source",
+      "url": "https://github.com/cleanlab/cleanlab"
+    },
     {
       "name": "Garak",
       "type": "open-source",