#!/usr/bin/env node
// msd-hook-version: {{MSD_VERSION}}
// MSD Read Injection Scanner — PostToolUse hook (#2201)
// Pattern-based pre-filter / blocklist: scans content returned by Read, WebFetch,
// and WebSearch for known prompt-injection patterns (regex + heuristic rules).
// This is a static pattern match — NOT a semantic guard, NOT PromptArmor.
// It does NOT understand context, intent, or novel phrasing; it catches
// known injection signatures at ingestion before they enter conversation context.
//
// Defense-in-depth: long MSD sessions hit context compression, and the
// summariser does not distinguish user instructions from content read from
// external files. Poisoned instructions that survive compression become
// indistinguishable from trusted context. This hook warns at ingestion time.
// Prompt-level self-guard and task-anchor controls (untrusted-input-boundary.md)
// operate independently as a complementary layer.
//
// Triggers on: Read, WebFetch, WebSearch PostToolUse events
// Action: Advisory warning by default; blocks HIGH only when security.injection_blocking=true
// Severity: LOW (1–2 patterns), HIGH (3+ patterns)
//
// False-positive exclusion: .planning/, REVIEW.md, CHECKPOINT, security docs,
// hook source files — these legitimately contain injection-like strings.

const path = require('path');
const fs = require('fs');
const { HOOK_ON_CRASH, allow, crash } = require('./lib/hook-exit.js');

// This is a PostToolUse advisory scanner over content the tool call already
// returned; a crash while scanning must not retroactively block that Read/
// WebFetch/WebSearch result from reaching the agent — losing the injection
// check is safer than failing the tool call it is only observing (#3911).
const ON_CRASH = HOOK_ON_CRASH.ALLOW;

// Summarisation-specific patterns (novel — not in msd-prompt-guard.js).
// These target instructions specifically designed to survive context compression.
const SUMMARISATION_PATTERNS = [
  /when\s+(?:summari[sz]ing|compressing|compacting),?\s+(?:retain|preserve|keep)\s+(?:this|these)/i,
  /this\s+(?:instruction|directive|rule)\s+is\s+(?:permanent|persistent|immutable)/i,
  /preserve\s+(?:these|this)\s+(?:rules?|instructions?|directives?)\s+(?:in|through|after|during)/i,
  /(?:retain|keep)\s+(?:this|these)\s+(?:in|through|after)\s+(?:summar|compress|compact)/i,
];

// Markdown link patterns — mirrors scripts/security.cjs MARKDOWN_LINK_PATTERNS, inlined for hook independence.
// Issue #113: detect javascript:, data: (non-safe-list), userinfo credentials, and token-in-query.
//
// Sources:
//   MD-LINK-JS-SCHEME: OWASP XSS Prevention
//     https://cheatsheetseries.owasp.org/cheatsheets/Cross_Site_Scripting_Prevention_Cheat_Sheet.html
//   MD-LINK-DATA-SCHEME: OWASP File Upload (SVG unsafe)
//     https://cheatsheetseries.owasp.org/cheatsheets/File_Upload_Cheat_Sheet.html#svg-files
//   MD-LINK-USERINFO: RFC 3986 §3.2.1, RFC 9110 §4.2.4
//     https://www.rfc-editor.org/rfc/rfc3986#section-3.2.1
//     https://www.rfc-editor.org/rfc/rfc9110#section-4.2.4
//   MD-LINK-TOKEN-IN-QUERY: RFC 9700 §4.3.1
//     https://www.rfc-editor.org/rfc/rfc9700#section-4.3.1
const DATA_URI_SAFE_MIME_RE = /^data:(image\/(png|jpe?g|gif|webp|bmp|ico|avif|heic)|font\/(woff2?|otf|ttf))(;[^,]*)?,/i;

const MARKDOWN_LINK_PATTERNS = [
  {
    pattern: /\]\(\s*javascript:/i,
    ruleId: 'MD-LINK-JS-SCHEME',
  },
  {
    pattern: /\]\(\s*data:/i,
    ruleId: 'MD-LINK-DATA-SCHEME',
    safePredicate: (line) => {
      const m = line.match(/\]\(\s*(data:[^)]*)/i);
      if (!m) return false;
      return DATA_URI_SAFE_MIME_RE.test(m[1]);
    },
  },
  {
    pattern: /\]\(\s*https?:\/\/[^/\s]+:[^/@\s]+@/i,
    ruleId: 'MD-LINK-USERINFO',
  },
  {
    pattern: /[?&](token|access_token|id_token|refresh_token|api_key|apikey|secret|password|client_secret|code)=/i,
    ruleId: 'MD-LINK-TOKEN-IN-QUERY',
  },
];

// Standard injection patterns — shared with msd-prompt-guard.js via
// hooks/lib/injection-patterns.js so the two surfaces cannot drift (#3504).
// Staging of the lib helper is allowlisted in MSD_HOOK_LIB_FILES (bin/install.js).
const { INJECTION_PATTERNS, describePattern } = require('./lib/injection-patterns.js');

const ALL_PATTERNS = [...INJECTION_PATTERNS, ...SUMMARISATION_PATTERNS];

// #3023: the staged bundle's directory name is runtime-descriptor-driven, so a
// literal `/<config>/hooks/` fragment cannot reliably identify MSD's own hook
// scripts. This module lives inside the bundle, so __dirname identifies it by
// construction. Normalized to forward slashes to match `p` below.
const OWN_BUNDLE_PREFIX = __dirname.replace(/\\/g, '/').replace(/\/+$/, '') + '/';

// Synthetic rule ids for the finding classes that have no entry in
// MARKDOWN_LINK_PATTERNS. Frozen and referenced from BOTH the push sites and
// renderFinding so the two can never drift — a bare literal repeated at each
// site is how a rename silently falls through to the generic render branch.
const RULE_IDS = Object.freeze({
  INJECTION_PATTERN: 'INJECTION-PATTERN',
  INVISIBLE_UNICODE: 'INVISIBLE-UNICODE',
  UNICODE_TAG_BLOCK: 'UNICODE-TAG-BLOCK',
});

function isExcludedPath(filePath) {
  const p = filePath.replace(/\\/g, '/');
  return (
    p.includes('/.planning/') ||
    p.includes('.planning/') ||
    /(?:^|\/)REVIEW\.md$/i.test(p) ||
    /CHECKPOINT/i.test(path.basename(p)) ||
    /[/\\](?:security|techsec|injection)[/\\.]/i.test(p) ||
    /security\.cjs$/.test(p) ||
    p.startsWith(OWN_BUNDLE_PREFIX) ||
    p.includes('/.claude/hooks/')
  );
}

let inputBuf = '';
const stdinTimeout = setTimeout(() => allow(undefined), 5000);
process.stdin.setEncoding('utf8');
process.stdin.on('data', chunk => { inputBuf += chunk; });
process.stdin.on('end', () => {
  clearTimeout(stdinTimeout);
  try {
    const data = JSON.parse(inputBuf);

    const toolName = data.tool_name;
    const SCANNED_TOOLS = new Set(['Read', 'WebFetch', 'WebSearch']);
    if (!SCANNED_TOOLS.has(toolName)) {
      allow(undefined);
    }

    // Source label + path-exclusion (path-exclusion applies to file reads only)
    let source;
    if (toolName === 'Read') {
      // #2595 (review Major 3, sibling sweep): typed read — a non-string
      // threw inside isExcludedPath()'s .replace() into the outer catch.
      source = typeof data.tool_input?.file_path === 'string'
        ? data.tool_input.file_path
        : '';
      if (!source) allow(undefined);
      if (isExcludedPath(source)) allow(undefined);
    } else if (toolName === 'WebFetch') {
      source = data.tool_input?.url || 'web';
    } else { // WebSearch
      source = `search: ${data.tool_input?.query || ''}`;
    }

    // Extract content from tool_response — string, {content}, or arbitrary object
    let content = '';
    const resp = data.tool_response;
    if (typeof resp === 'string') {
      content = resp;
    } else if (resp && typeof resp === 'object') {
      const c = resp.content;
      if (Array.isArray(c)) {
        content = c.map(b => (typeof b === 'string' ? b : b.text || '')).join('\n');
      } else if (c != null) {
        content = String(c);
      } else {
        // WebSearch results etc. — scan the serialized response
        try { content = JSON.stringify(resp); } catch { content = ''; }
      }
    }

    if (!content || content.length < 20) {
      allow(undefined);
    }

    // Typed findings IR — single source of truth for both the machine-readable
    // `findings` array and the rendered advisory prose. Never build these as two
    // parallel arrays: that invites the generative-fix-divergence defect class
    // where the rendered text and the structured data silently drift apart.
    const findings = [];

    for (const pattern of ALL_PATTERNS) {
      if (pattern.test(content)) {
        // Trim pattern source for readable output (shared with msd-prompt-guard.js)
        findings.push({
          ruleId: RULE_IDS.INJECTION_PATTERN,
          match: describePattern(pattern),
        });
      }
    }

    // Markdown link patterns (issue #113)
    const lines = content.split('\n');
    for (const entry of MARKDOWN_LINK_PATTERNS) {
      for (let i = 0; i < lines.length; i++) {
        const line = lines[i];
        const m = line.match(entry.pattern);
        if (!m) continue;
        if (entry.safePredicate && entry.safePredicate(line)) continue;
        findings.push({ ruleId: entry.ruleId, match: m[0].substring(0, 40) });
      }
    }

    // Invisible Unicode (zero-width, RTL override, soft hyphen, BOM)
    if (/[\u200B-\u200F\u2028-\u202F\uFEFF\u00AD\u2060-\u2069]/.test(content)) {
      findings.push({ ruleId: RULE_IDS.INVISIBLE_UNICODE, match: null });
    }

    // Unicode tag block U+E0000–E007F (invisible instruction injection vector)
    try {
      if (/[\u{E0000}-\u{E007F}]/u.test(content)) {
        findings.push({ ruleId: RULE_IDS.UNICODE_TAG_BLOCK, match: null });
      }
    } catch {
      // Engine does not support Unicode property escapes — skip this check
    }

    if (findings.length === 0) {
      allow(undefined);
    }

    // Renders one finding back into the exact prose fragment the advisory has
    // always embedded. Kept as the ONLY place that maps IR -> text, so the
    // `additionalContext` string and the `findings` array can never diverge.
    function renderFinding(f) {
      if (f.ruleId === RULE_IDS.INVISIBLE_UNICODE) return 'invisible-unicode';
      if (f.ruleId === RULE_IDS.UNICODE_TAG_BLOCK) return 'unicode-tag-block';
      if (f.ruleId === RULE_IDS.INJECTION_PATTERN) return f.match;
      return `${f.ruleId}:${f.match}`;
    }

    const severity = findings.length >= 3 ? 'HIGH' : 'LOW';
    const label = toolName === 'Read' ? path.basename(source) : source;
    const detail = severity === 'HIGH'
      ? 'Multiple patterns — strong injection signal. Review for embedded instructions before proceeding.'
      : 'Single pattern match may be a false positive (e.g., documentation). Proceed with awareness.';
    const advisory =
      `\u26a0\ufe0f INJECTION SCAN [${severity}] (${toolName}): "${label}" triggered ` +
      `${findings.length} pattern(s): ${findings.map(renderFinding).join(', ')}. ` +
      `This content is now in your conversation context. ${detail} Source: ${source}`;

    // Opt-in blocking: only when configured AND high-confidence
    let blocking = false;
    if (severity === 'HIGH') {
      try {
        const cfgBase = data.cwd || process.cwd();
        const cfgPath = path.join(cfgBase, '.planning', 'config.json');
        const cfg = JSON.parse(fs.readFileSync(cfgPath, 'utf8'));
        blocking = cfg.security?.injection_blocking === true;
      } catch { /* no config ⇒ advisory */ }
    }

    const output = blocking
      ? { decision: 'block',
          reason: `Prompt-injection blocked (${toolName}). ${advisory}`,
          hookSpecificOutput: { hookEventName: 'PostToolUse', additionalContext: advisory, findings, severity, source } }
      : { hookSpecificOutput: { hookEventName: 'PostToolUse', additionalContext: advisory, findings, severity, source } };

    process.stdout.write(JSON.stringify(output));
  } catch {
    // Silent fail — never block tool execution.
    // ON_CRASH is declared ALLOW at module top: this preserves today's
    // exit(0) fail-open behavior exactly (#3911).
    crash(ON_CRASH, undefined);
  }
});