ClawHub
Back to skill

Security audit

Luca Assistant

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed credit-card rewards assistant that installs an external package and keeps a local portfolio database, with some supply-chain and consent cautions but no artifact-backed malicious behavior.

Install only if you trust the luca-assistant PyPI package and publisher. Treat portfolio write actions as persistent local changes, and have the agent confirm before adding, closing, importing, or updating card data.

A.I.G

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (1)

T08 · Insecure Dependencies

Warning
Location
scripts/setup.sh:16
Finding

Unpinned Third-Party Package Is Installed and Immediately Executed

Content
View full analysis

Vulnerability Details

File Location: scripts/setup.sh, lines 16-25
Vulnerability Type: Unpinned executable dependency and mutable supply-chain artifact
Risk Level: Medium

Vulnerable Code

bash
# 1. Install luca-assistant from PyPI
if command -v luca-mcp &>/dev/null; then
  echo "[ok] luca-assistant already installed"
else
  echo "[..] Installing luca-assistant from PyPI..."
  uv tool install luca-assistant
  echo "[ok] Installed"
fi

# 2. Create data directory and initialize the database
mkdir -p "${LUCA_DATA_DIR}"
echo "[..] Initializing database (downloading Offer Optimist card data)..."
luca init

Technical Analysis

The setup script installs luca-assistant from PyPI without specifying an audited version, validating a package hash, checking a signature, or using a locked dependency manifest. Consequently, the code installed by the same reviewed skill can change between executions.

The script then invokes the installed luca executable through luca init. The dependency's source code is not included in this project, so its effective behavior—including initialization and network activity—cannot be verified from the audited files. The homepage declared in SKILL.md does not cryptographically establish that the retrieved PyPI artifact corresponds to a reviewed source revision.

This is a supply-chain weakness rather than evidence that the current upstream package is malicious.

Attack Path

  1. An attacker compromises the upstream PyPI publisher account, publishing pipeline, package distribution process, or a future release.
  2. The attacker publishes a modified luca-assistant package under the expected package name.
  3. A user runs scripts/setup.sh on a system where luca-mcp is not already installed.
  4. uv tool install luca-assistant resolves and downloads the attacker-controlled release because no version or artifact hash is pinned.
  5. The setup script invokes luca init, causing code from the newly installed pa ...[truncated 691 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin luca-assistant to a specific version that has undergone security review, rather than resolving the latest release.
  2. Verify the exact distribution artifact using a trusted cryptographic hash or package signature.
  3. Maintain a reviewed lockfile or equivalent reproducible dependency manifest, including transitive dependencies.
  4. Retrieve packages only from a trusted, explicitly configured package index.
  5. Separate installation from execution so users can verify the installed artifact before luca init runs.
  6. Run initialization with least privilege and a restricted environment containing only required variables.
  7. Document the expected network endpoints and downloaded data used by luca init.
  8. Establish a controlled update process that reviews upstream changes before modifying the pinned version or hashes.

SkillSpector

Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill invokes shell commands and installs/executes network-backed packages, but it does not declare any explicit tool scope such as allowed tools or permissions. That creates an authorization gap: an agent may be induced to perform package installation, shell execution, and MCP invocation without clear confinement, increasing the chance of unintended command execution or data access.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill exposes state-changing portfolio operations such as adding or closing cards, which modify stored personal financial metadata, but it provides no warning or consent guidance before those actions. In an agent setting, this can lead to silent or mistaken modification of a user's long-lived portfolio data, causing privacy issues, inaccurate financial tracking, or downstream bad recommendations.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · scripts/setup.sh (reported line 23)May include surrounding context.

sh
echo "[ok] Installed"
fi

# 2. Create data directory and initialize the database
mkdir -p "${LUCA_DATA_DIR}"
echo "[..] Initializing database (downloading Offer Optimist card data)..."
luca init

Static analysis

No suspicious patterns detected.