open/ktg-plugin-marketplace
1
0
Fork 0
Code Issues Pull requests Releases Wiki Activity Actions

chore(llm-security): v7.7.2 — language consistency pass

Browse source 
~/.claude/CLAUDE.md specifies English for code and documentation,
Norwegian for dialog only. Norwegian had crept into surface text
across v7.5-v7.7. Translated to English in eight surfaces.

No scanner, hook, or behavior changes — purely surface text.

- 18 skill commands: the HTML Report-step now reads "HTML report:
  [Open in browser]" instead of "HTML-rapport: [Åpne i nettleser]"
- scripts/lib/report-renderers.mjs: key-stat labels, lede defaults,
  table headers, maturity-ladder descriptions, action-tier labels,
  clean buckets, dry-run/apply copy, and JS comments. Regex
  alternations /^high|^høy/ and /resolution|løsning/i preserved.
- playground/llm-security-playground.html: same renderer changes
  mirrored bit-identical, plus playground-only UI strings (catalog,
  breadcrumb aria-label, theme toggle, builder-modal hint,
  guide-panel "no projects yet", delete confirmation, alert/copy).
  Demo-state fixture content for dft-komplett-demo preserved
  (intentional Norwegian persona).
- agents/skill-scanner-agent.md + agents/mcp-scanner-agent.md:
  Generaliseringsgrense + Parallell Read-strategi sections translated
  to Generalization boundary + Parallel Read strategy.
- README.md: playground architecture prose + Recent versions table
  (v7.5.0 — v7.7.1).
- CLAUDE.md: v7.7.1 highlights translated, new v7.7.2 highlights
  added.
- ../../README.md: llm-security v7.5.0 — v7.7.1 bullets.
- ../../CLAUDE.md: llm-security catalog entry.
- docs/scanner-reference.md: six runnable-examples table cells.
- docs/version-history.md: new v7.7.2 entry. v7.5-v7.7 narrative
  sections left in original language (deferred per operator).
- Version bumped 7.7.1 → 7.7.2 in package.json,
  .claude-plugin/plugin.json, README badge + Recent versions,
  CLAUDE.md header + state, docs/version-history.md, playground
  renderHome hardcoded string, root README + CLAUDE.md llm-security
  entries.

Tests: 1820/1820 green. CLI smoke-test: 18/18 commandIds produce
>138 KB self-contained HTML. Browser-dogfood verified.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
Kjell Tore Guttormsen 2026-05-19 06:47:44 +02:00
commit 03b8885b6e
31 changed files with 467 additions and 359 deletions
Download patch file Download diff file Expand all files Collapse all files

24
plugins/llm-security/docs/scanner-reference.md
View file

@ -105,18 +105,18 @@ Scan reports are stored in `reports/` as `.docx` (for sharing) with `.md` source
## Examples (runnable demonstrations)
Self-contained, deterministic threat-fixture mappes under `examples/`. Each mappe har `README.md`, fixture/script/transcript, `run-*.{sh,mjs}`, og `expected-findings.md`. Demonstrasjoner — ikke unit-tester.
Self-contained, deterministic threat fixtures live under `examples/`. Each directory has a `README.md`, fixture/script/transcript, `run-*.{sh,mjs}`, and `expected-findings.md`. They are demonstrations — not unit tests.
| Mappe | Demonstrerer | Hooks/scanners | Sentinel |
| Directory | Demonstrates | Hooks/scanners | Sentinel |
|-------|--------------|----------------|----------|
| `malicious-skill-demo/` | Skill scanner end-to-end (UNI/ENT/PRM/DEP/TNT/NET + 7 LLM-kategorier) | `scan-orchestrator` + agents | BLOCK 100/100 |
| `prompt-injection-showcase/` | 61 payloads × 19 kategorier mot `pre-prompt-inject-scan`, `post-mcp-verify`, `pre-bash-destructive` | runtime hooks | per-kategori expected outcome |
| `lethal-trifecta-walkthrough/` | Rule-of-Two advisory på leg 3 (WebFetch → Read .env → Bash curl POST) + suppression | `post-session-guard` | advisory på stage 3 |
| `mcp-rug-pull/` | Cumulative drift-advisory (E14, v7.3.0) — 7 stadier under per-update-terskel, kumulativt over 25% baseline | `post-mcp-verify` + `mcp-description-cache.mjs` | advisory på stage 7 |
| `supply-chain-attack/` | PreToolUse-blokk på kompromittert pakke + scope-hop advisory + dep-auditor typosquats + postinstall curl-pipe | `pre-install-supply-chain` + `dep-auditor` + `supply-chain-data` | 6+ funn, 2 advisories, 1 BLOCK |
| `poisoned-claude-md/` | 6 detektorer (injection / shell / URL / credential paths / permission expansion / encoded payloads) inkl. E15 agent-fil-overflate | `memory-poisoning-scanner` | ≥18 funn fordelt på 2 filer |
| `bash-evasion-gallery/` | T1-T9 disguised destructive commands → normalisert + blokkert (defense-in-depth over Claude Code 2.1.98+) | `pre-bash-destructive` + `bash-normalize` | 10 BLOCK eksitkoder |
| `toxic-agent-demo/` | Single-component lethal trifecta — agent med [Bash, Read, WebFetch] uten hook-guards = CRITICAL TFA-finding | `toxic-flow-analyzer` (TFA) | 1 CRITICAL `Lethal trifecta:` |
| `pre-compact-poisoning/` | PreCompact-hook fanger injection + AWS-shaped credential i syntetisk transcript på tvers av off/warn/block-modus | `pre-compact-scan` | 9 pass: block exit 2 + reason; warn systemMessage; off skip; benign passes |
| `malicious-skill-demo/` | Skill scanner end-to-end (UNI/ENT/PRM/DEP/TNT/NET + 7 LLM categories) | `scan-orchestrator` + agents | BLOCK 100/100 |
| `prompt-injection-showcase/` | 61 payloads × 19 categories against `pre-prompt-inject-scan`, `post-mcp-verify`, `pre-bash-destructive` | runtime hooks | per-category expected outcome |
| `lethal-trifecta-walkthrough/` | Rule-of-Two advisory on leg 3 (WebFetch → Read .env → Bash curl POST) + suppression | `post-session-guard` | advisory at stage 3 |
| `mcp-rug-pull/` | Cumulative drift advisory (E14, v7.3.0) — 7 stages below the per-update threshold, cumulatively over a 25% baseline shift | `post-mcp-verify` + `mcp-description-cache.mjs` | advisory at stage 7 |
| `supply-chain-attack/` | PreToolUse block on a compromised package + scope-hop advisory + dep-auditor typosquats + postinstall curl-pipe | `pre-install-supply-chain` + `dep-auditor` + `supply-chain-data` | 6+ findings, 2 advisories, 1 BLOCK |
| `poisoned-claude-md/` | 6 detectors (injection / shell / URL / credential paths / permission expansion / encoded payloads) including the E15 agent-file surface | `memory-poisoning-scanner` | ≥18 findings split across 2 files |
| `bash-evasion-gallery/` | T1-T9 disguised destructive commands → normalized + blocked (defense-in-depth over Claude Code 2.1.98+) | `pre-bash-destructive` + `bash-normalize` | 10 BLOCK exit codes |
| `toxic-agent-demo/` | Single-component lethal trifecta — an agent with [Bash, Read, WebFetch] and no hook guards = CRITICAL TFA finding | `toxic-flow-analyzer` (TFA) | 1 CRITICAL `Lethal trifecta:` |
| `pre-compact-poisoning/` | The PreCompact hook catches injection + an AWS-shaped credential in a synthetic transcript across off/warn/block modes | `pre-compact-scan` | 9 cases: block exit 2 + reason; warn systemMessage; off skip; benign passes |
State-isolering: alle eksempler som muterer global state bruker run-script PID (post-session-guard via `${ppid}.jsonl`) eller env-overrides (`LLM_SECURITY_MCP_CACHE_FILE` for MCP-cache). Brukerens reelle `/tmp/llm-security-session-*.jsonl` og `~/.cache/llm-security/` røres aldri.
State isolation: every example that mutates global state uses the run-script PID (post-session-guard via `${ppid}.jsonl`) or env overrides (`LLM_SECURITY_MCP_CACHE_FILE` for the MCP cache). The user's real `/tmp/llm-security-session-*.jsonl` and `~/.cache/llm-security/` are never touched.