Back to skill

Security audit

psyseal-scan

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because scanners found weaknesses in its claimed filesystem boundaries and unpinned npm commands, while inspection was incomplete.

Consider avoiding installation unless you trust the publisher and are comfortable with the unpinned npm execution and reported symlink boundary weaknesses. The scan is partial: only 12 of 32 components were fully inspected, and the tool-misuse analyzer was degraded, so the results cannot establish that unreported behavior is safe.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (23)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied command does not enforce offline execution and can fetch executable code from npm. These behaviors conflict with the explicit offline, read-only, and dependency-free description. Although a cached package could run without downloading, that is not guaranteed by this invocation. The omitted package implementation prevents verification of the advertised scanning behavior.

Content

No source excerpt is available for this finding.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/i18n.mjs (reported line 144)May include surrounding context.

js
f: {
      'secret.exposed': (items) => ({
        head: `${items.length} credencial(es) en texto plano en ${list(files(items), 4)}: ${list(items.map((i) => `${i.data.file}:${i.data.line} (${i.data.type})`), 6)}`,
        fix: 'Rota AHORA, en el proveedor, las claves marcadas (si estuvieron en un archivo que el agente lee, asúmelas comprometidas), borra esas líneas y guarda las claves donde tu plataforma guarda credenciales (.env o variables de entorno).',
      }),
      'config.secret_plaintext': (items) => ({
        head: `Credencial(es) en texto plano dentro de la configuración: ${list(items.map((i) => `${i.data.path} (${i.data.type})`), 6)}`,

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/i18n.mjs (reported line 343)May include surrounding context.

js
f: {
      'secret.exposed': (items) => ({
        head: `${items.length} plaintext credential(s) in ${list(files(items), 4)}: ${list(items.map((i) => `${i.data.file}:${i.data.line} (${i.data.type})`), 6)}`,
        fix: 'Rotate NOW, at the provider, the flagged keys (if they sat in a file your agent reads, assume they are compromised), delete those lines and keep keys where your platform stores credentials (.env or environment variables).',
      }),
      'config.secret_plaintext': (items) => ({
        head: `Plaintext credential(s) inside the config: ${list(items.map((i) => `${i.data.path} (${i.data.type})`), 6)}`,

Memory Manipulation

High
Category
Memory Poisoning
Confidence
90% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · src/i18n.mjs (reported line 355)May include surrounding context.

js
}),
      'inject.override': (items) => ({
        head: `Text that tries to rewrite the agent rules in ${list(items.map((i) => `${i.data.file}:${i.data.line}`), 6)}`,
        fix: 'Review the flagged text: if you wrote it, keep it; if it came from external content, delete it (this is the classic way to poison memory).',
      }),
      'inject.hidden': (items) => ({
        head: `Invisible characters (used to hide instructions) in ${list(items.map((i) => `${i.data.file}:${i.data.line}`), 6)}`,

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/platforms/generic.mjs (reported line 52)May include surrounding context.

js
daily,
        schedules: [],
        symlinksSkipped: list.symlinks,
        stores: secretStores(root, ['.env', '.env.local']),
        notes: [],
        generic: true,
        listedTruncated: list.truncated,

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/platforms/hermes.mjs (reported line 19)May include surrounding context.

js
export function detect({ home, env }) {
  const root = env.HERMES_HOME ? path.resolve(expandHome(env.HERMES_HOME, home)) : path.join(home, '.hermes');
  if (!isDir(root)) return [];
  const looksLike = ['config.yaml', 'memories', 'SOUL.md', 'skills', 'state.db', '.env'].some((n) => fs.existsSync(path.join(root, n)));
  return looksLike ? [{ platform: ID, root }] : [];
}

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/platforms/zeroclaw.mjs (reported line 96)May include surrounding context.

js
const snapshot = fileEntry(path.join(dataDir, 'memory', 'MEMORY_SNAPSHOT.md'), 'data/memory/MEMORY_SNAPSHOT.md', 'memory', { agentWritten: true });
  if (snapshot.present && agents.length) agents[0].files.push(snapshot);

  const stores = secretStores(root, ['config.toml', '.secret_key', '.env']);
  if (agents.length) agents[0].stores = stores;
  return { platform: ID, name: NAME, root, agents, stores: agents.length ? [] : stores };
}

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/platforms/generic.mjs (reported line 52)May include surrounding context.

js
daily,
        schedules: [],
        symlinksSkipped: list.symlinks,
        stores: secretStores(root, ['.env', '.env.local']),
        notes: [],
        generic: true,
        listedTruncated: list.truncated,

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/platforms/hermes.mjs (reported line 97)May include surrounding context.

js
daily: null,
    schedules,
    symlinksSkipped: skillList.symlinks + files.filter((f) => f.symlink).length,
    stores: secretStores(h, ['.env', 'auth.json', '.anthropic_oauth.json']),
    notes: jobsError ? [{ id: 'schedule.unreadable', data: { file: 'cron/jobs.json' } }] : [],
  };
}

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/platforms/nanobot.mjs (reported line 130)May include surrounding context.

js
daily: null,
    schedules,
    symlinksSkipped: skillList.symlinks + files.filter((f) => f.symlink).length,
    stores: secretStores(root, ['config.json', '.env']),
    notes: [],
  };
  return { platform: ID, name: NAME, root, agents: [agent] };

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/platforms/nanoclaw.mjs (reported line 75)May include surrounding context.

js
notes: [],
    });
  }
  const stores = secretStores(root, ['.env']);
  if (agents.length) agents[0].stores = stores;
  return { platform: ID, name: NAME, root, agents, stores: agents.length ? [] : stores, scheduleNote: 'sqlite' };
}

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · src/platforms/nanoclaw.mjs (reported line 3)May include surrounding context.

js
// NanoClaw. Se instala clonando su repositorio (la carpeta la elige la persona). Cada grupo de agente vive en
// groups/<carpeta>/ con su CLAUDE.md (instrucciones y memoria, entra completo al contexto) y sus skills;
// las credenciales están en .env en la raíz de la instalación.

import fs from 'node:fs';
import path from 'node:path';

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · src/scan.mjs (reported line 232)May include surrounding context.

js
function injectionFindings(findings, file, text, strict, { skill = false, abs = null, fileKind = null } = {}) {
  for (const s of scanInjection(text, { strict })) {
    // En skills son comunes las instrucciones de instalación (curl ... | sh): se marcan para revisar, no como alarma.
    const severity = s.cls === 'critical' && !skill ? 'high' : 'medium';
    findings.push({ id: `inject.${s.cls}`, severity, data: { file, abs, fileKind: skill ? 'skill' : fileKind, strict, line: s.line, signal: s.id, count: s.count } });
  }

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · src/scan.mjs (reported line 232)May include surrounding context.

js
function injectionFindings(findings, file, text, strict, { skill = false, abs = null, fileKind = null } = {}) {
  for (const s of scanInjection(text, { strict })) {
    // En skills son comunes las instrucciones de instalación (curl … | sh): se marcan para revisar, no como alarma.
    const severity = s.cls === 'critical' && !skill ? 'high' : 'medium';
    findings.push({ id: `inject.${s.cls}`, severity, data: { file, abs, fileKind: skill ? 'skill' : fileKind, strict, line: s.line, signal: s.id, count: s.count } });
  }

MCP server referenced without pinned version: 'npx psyseal-scan'.

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The documented command executes an unpinned npm package, so the code obtained can change without review. A compromised future release could execute with the user's privileges and access the sensitive agent files this scanner is intended to inspect; this is a conditional supply-chain risk, not evidence of current malicious behavior.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The installation example executes the unpinned skills CLI and obtains skill content from an unpinned repository reference. Although this is an installation command rather than an MCP server invocation, changing or compromised upstream code can still execute locally or introduce malicious agent instructions.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y psyseal-scan'.

Medium
Category
MCP Rug Pull
Confidence
98% confidence
Finding

The MCP launch command uses an unpinned package with automatic installation approval. On launch, npm may obtain a different release, creating a supply-chain execution risk in a process that can read sensitive local agent data.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y psyseal-scan'.

Medium
Category
MCP Rug Pull
Confidence
98% confidence
Finding

The Spanish MCP example automatically executes an unpinned package. A compromised or unexpected release could run with the user's privileges; automatic approval removes the package-install confirmation step.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx psyseal-scan'.

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The Spanish quick-start repeats the unpinned npm execution command. Users following it may execute an unreviewed future release with access to local files, despite the scanner's privacy-oriented purpose.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The --usage file reader follows symlinks despite the skill's explicit no-symlink guarantee.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

Both statSync and readFileSync follow symbolic links, allowing an attacker-controlled usage-file path to redirect reads outside the intended location. This violates the explicit no-symlink guarantee, although exploitation requires this optional input and the shown code does not directly print arbitrary file contents.

Content

No source excerpt is available for this finding.

Configuration loading follows symlinks, contradicting the skill's declared filesystem boundary.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The configuration reader follows symlinks for both the initial configuration and included files, contrary to the declared boundary. Include realpath checks constrain stable include targets to allowed roots, but do not reject symlinks, protect the initial configuration, or prevent replacement between validation and reading; substituted configuration can also redirect subsequent workspace selection.

Content

No source excerpt is available for this finding.

The documented no-symlink policy is not enforced for parent directories.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

lstatSync rejects a symlink only at the final component; the operating system still follows symlinked parent directories. An attacker who controls a parent component can redirect scanning outside the intended tree, and the separate check and read also permit replacement races, undermining the advertised filesystem boundary.

Content

No source excerpt is available for this finding.