registry / claude-recall / 0.44.0

claude-recall@0.44.0

Persistent memory for Claude Code and Pi with native Skills integration, automatic capture, failure learning, and project scoping

AI Security Review

scanned 10d ago · by lpm-firewall-ai

The install hook performs an unprompted, home-wide repair of package-owned Claude Code hooks, creating a real agent configuration lifecycle risk. The repair is narrowly scoped and backed up, with no inspected evidence of unrelated configuration changes or a concrete attack chain.

Static reason

High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review

Trigger

Installing the package through npm triggers its postinstall script.

Impact

Existing agent settings may be changed during package installation, enabling the repaired Claude Recall hooks to run from the current package location.

Mechanism

Postinstall runs repair with automatic mode and home-wide scope. Repair scans user and project settings, then backs up and rewrites a recognized Claude Recall hook command when its absolute path is stale and a replacement can be resolved.

AI rationale

Evidence

package.jsonscripts/postinstall.jsdist/cli/commands/repair.js~/.claude/settings.jsonproject .claude/settings.json files

Decision evidence

public snapshot

AI called this Suspicious at 91.0% confidence as Dangerous Capability with medium false-positive risk.

Evidence for warning

`package.json` runs `scripts/postinstall.js` during installation; that script automatically launches repair with `--auto --scope all`.
The repair scans Claude Code settings at user scope and across projects under the home directory. It targets recognizable Claude Recall hooks and proposes a replacement only for stale commands when the hook ID is known and the package is on `PATH`.
When a repair applies, it backs up each settings file before writing the change. The install hook supplies `--auto`, so this repair can happen without an interactive confirmation.
The other inspected integration code shows explicit CLI setup for installing hooks, project skill generation, and memory sync to Claude Code’s memory directory. The MCP repair module addresses server registration separately.

Evidence against

The automatic repair is limited to existing Claude Recall hook commands; the inspected logic does not show it adding hooks or changing unrelated hook commands.

Behavioral surface

Source

ChildProcessCryptoEnvironmentVars

Source & flagged code

20 flagged

package.jsonView file

Lines 13-53json

13 "skills/",

14 "scripts/postinstall.js",

15 "scripts/uninstall.js",

16 "README.md",

17 "LICENSE",

18 "docs/"

19 ],

20 "pi": {

21 "extensions": [

22 "./dist/pi/extension.js"

23 ],

24 "skills": [

25 "./skills/*.md"

26 ]

27 },

28 "directories": {

29 "doc": "docs",

30 "test": "tests"

31 },

32 "scripts": {

33 "postinstall": "node scripts/postinstall.js || true",

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L33

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg · L33

34 "test": "jest",

35 "test:watch": "jest --watch",

36 "test:unit": "jest tests/unit",

37 "test:integration": "jest tests/integration",

38 "test:mcp": "jest tests/mcp",

39 "test:watch:unit": "jest --watch tests/unit",

40 "test:coverage": "jest --coverage",

41 "test:coverage:unit": "jest --coverage --maxWorkers=50% tests/unit",

42 "test:coverage:integration": "jest --coverage --maxWorkers=50% tests/integration",

43 "test:benchmarks": "jest tests/benchmarks",

44 "bench:retrieval": "ts-node tests/benchmarks/retrieval-benchmark.ts",

45 "test:experimental": "jest tests/experimental",

46 "build": "tsc && mkdir -p dist/memory dist/config && cp src/memory/schema.sql dist/memory/ && (cp -r src/config/* dist/config/ 2>/dev/null || true) && chmod +x dist/cli/claude-recall-cli.js",

47 "build:cli": "tsc && chmod +x dist/cli/claude-recall-cli.js",

48 "dev": "ts-node",

49 "start": "ts-node src/server.ts",

50 "start:dev": "nodemon --exec ts-node src/server.ts",

51 "claude-recall": "ts-node src/cli/claude-recall-cli.ts",

52 "install:global": "npm run build && npm link",

53 "stats": "npm run claude-recall stats",

scripts/postinstall.jsView file

Lines 1-33javascript

1#!/usr/bin/env node

3const fs = require('fs');

4const path = require('path');

5const os

dist/cli/claude-recall-cli.jsView file

•path = dist/cli/claude-recall-cli.js kind = trigger_reachable_excluded_payload sizeBytes = 115749 triggerReachableSource = dist/cli/claude-recall-cli.js excludedPayload = .claude/hooks/search_enforcer.py

Critical

Trigger Reachable Excluded Directory Payload

A package entrypoint or install-reachable source explicitly loads and activates an executable-looking payload from a test, fixture, or hidden path.

dist/cli/claude-recall-cli.jsView on unpkg

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. process.exit(1); } } truncatecontent(content) { const str = typeof content === 'string' ? content : json.stringify(content); const maxlength = 100; if (str.length <= maxlength) return str; return str.substring(0, maxlength) + '...'; } } function getversion() { try { const packagejson = json.parse(fs.readfilesync(path.join(__dirname, '../../package.json'), 'utf-8')); return packagejson.version; } catch (error) { return '0.3.0'; } } async function main() { program .name('claude-recall') .descr

High

Trigger Reachable External Ai Agent Control Surface Mutation

Manifest-trigger-reachable source writes behavior-bearing configuration into a user or project AI-agent control surface.

dist/cli/claude-recall-cli.jsView on unpkg

•matchType = token_shingles matchedPackage = claude-recall@0.38.0 matchedPath = dist/cli/claude-recall-cli.js matchedIdentity = npm:Y2xhdWRlLXJlY2FsbA:0.38.0 similarity = 1.000 shingleOverlap = 48 summary = source token shingles overlapped finalized malicious source

.claude/hooks/search_enforcer.pyView file

•path = .claude/hooks/search_enforcer.py kind = payload_in_excluded_dir sizeBytes = 10876 magicHex = [redacted]

High

Payload In Excluded Dir

Package hides binary, compressed, or executable-looking payloads in test/fixture/hidden paths.

.claude/hooks/search_enforcer.pyView on unpkg

•path = .claude/hooks/search_enforcer.py kind = build_helper sizeBytes = 10876 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

.claude/hooks/search_enforcer.pyView on unpkg

dist/cli/commands/hook-commands.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 5

High

Semantic Analysis Limited

A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.

dist/cli/commands/hook-commands.js#virtual:normalized:round1View on unpkg

dist/hooks/kiro-hooks.jsView file

•matchType = normalized_sha256 matchedPackage = claude-recall@0.38.0 matchedPath = dist/hooks/kiro-hooks.js matchedIdentity = npm:Y2xhdWRlLXJlY2FsbA:0.38.0 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High