open/llm-security
1
0
Fork 0
Code Issues Releases Activity

docs(llm-security): v8 Phase 2 — B10 docs consistency, counts pinned by test

Browse source 
Extends tests/lib/doc-consistency.test.mjs with 15 cases that derive every
inventory count from source instead of trusting prose. Each count has one
stated derivation; a doc surface that disagrees now fails the suite.

Counts corrected (all were wrong before the test existed):
- orchestrated scanners: docs said 10 (README, ci-cd-guide), CLAUDE.md said 12,
  the synthesizer agent said 9 — scan-orchestrator registers 14
- total scanners: README badge + 3 prose sites said 23; the counting rule in
  docs/scanner-reference.md (14 orchestrated + 8 standalone) yields 22
- knowledge files: README badge + prose said 22; knowledge/ holds 23
- output.mjs finding() prefix JSDoc listed 10 of the 17 prefixes actually
  passed to it (missing IDE, MCI, MEM, PST, SCR, TFA, WFL)
- norwegian-context.md said "8 hooks, 10 scanners" -> 9 and 14
- ci-cd-guide "what gets scanned" table listed 10 of 14 rows; adds workflow,
  trigger abuse, signature, AST taint

Two plan items changed after verifying against ground truth:
- CLAUDE.md's synthesizer "(12 scanners)" was not a deliberate subset; the
  agent file itself claimed 9. Both bumped to 14.
- compliance-mapping.md's "13 posture categories" is substantively correct —
  its matrix has exactly 13 data rows, and categories 14-16 are governance
  consumers of the file, not rows in it. The planned 13->16 bump would have
  made the document false. Wording clarified to "code-level" instead, and the
  test now pins row count against the stated claim.

Framework currency (both verified against primary reporting):
- EU AI Act: Digital Omnibus (EP 2026-06-16, Council 2026-06-29) deferred the
  high-risk obligations behind Art. 9/15 to 2027-12-02 (Annex III) and
  2028-08-02 (Annex I); transparency still applies from 2026-08-02
- OWASP Agentic AI Top 10 labelled as the 2026 edition

Also: CLAUDE.md Distribution section rewritten monorepo -> polyrepo (each
plugin is its own repo; the catalog pins url + ref per plugin), and
current-state test counts synced 2013 -> 2034. Release-note paragraphs keep
their historical numbers.

No scanner, hook, or command behaviour changes. Suite 2034/2034.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wt4YQGoXwRja5K2Zmv8RZE
This commit is contained in:
Kjell Tore Guttormsen 2026-08-02 21:22:20 +02:00
commit 0f1be986d0
8 changed files with 287 additions and 22 deletions
Download patch file Download diff file Expand all files Collapse all files

4
knowledge/compliance-mapping.md
View file

@ -1,12 +1,12 @@
# Compliance Mapping
Maps the llm-security plugin's 13 posture categories and mitigation controls to three enterprise compliance frameworks: EU AI Act, NIST AI RMF, and ISO 42001.
Maps the llm-security plugin's 13 code-level posture categories and mitigation controls to three enterprise compliance frameworks: EU AI Act, NIST AI RMF, and ISO 42001.
Used by `posture-assessor-agent` and compliance-aware posture categories (14-16) to evaluate framework alignment.
## How to Read This Matrix
- **Plugin Control:** One of the 13 posture scanner categories
- **Plugin Control:** One of the 13 code-level posture categories (1-13). The three governance categories (14-16) are consumers of this matrix, not rows in it
- **Control Type:** Automated (hooks), Configured (settings), Advisory (scans/audits)
- **EU AI Act:** Regulation (EU) 2024/1689 article(s) the control satisfies
- **NIST AI RMF:** AI 100-1 function(s) the control supports (Govern, Map, Measure, Manage)

2
knowledge/norwegian-context.md
View file

@ -118,7 +118,7 @@ regulatory sandbox for controlled testing under the AI Act.
| Data protection in AI | GDPR, Datatilsynet sandbox | Secrets protection hooks, path guarding, credential scanning | Full |
| Transparency and explainability | Digdir principles, AI Act Art. 13 | Scan reports, posture reports, AI-BOM | Partial |
| Human oversight | Digdir principles, AI Act Art. 14 | Human Review Requirements (PST-07), Rule of Two, deny-first config | Full |
| Cybersecurity | AI Act Art. 15, NSM grunnprinsipper | All 8 hooks, 10 scanners, prompt injection hardening | Full |
| Cybersecurity | AI Act Art. 15, NSM grunnprinsipper | All 9 hooks, 14 scanners, prompt injection hardening | Full |
| Record-keeping | AI Act Art. 12, NSM detect principle | Audit trail (JSONL), session logging, baseline diffs | Full (v6.0) |
| Quality management | AI Act Art. 17 | Test suite (1147+ tests), posture scanner, scan-orchestrator | Partial |
| Supply chain integrity | AI Act Art. 15, NSM identify principle | Supply chain hooks, dep audit scanner, AI-BOM | Full |