Back to skill

Security audit

Monet Works Content Pack

Security checks for vulnerabilities and agentic risk

Overview

This is a small content-pack descriptor with manual install instructions for two named skills and no executable code, though users should verify the referenced packages before installing.

Before installing, check that `monet-works-content-qa-dv` and `ogilvy-humanizer-dv` come from the publisher you expect and review their own permissions and behavior. The pack itself does not include executable code or request sensitive access.

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
SKILL.md:24
Finding
Unpinned External Skill Dependencies## Vulnerability Details **File Location**: `SKILL.md`, lines 24–25 **Vulnerability Type**: Unpinned third-party skill installation **Risk Level**: Medium ```bash clawhub install monet-works-content-qa-dv clawhub install ogilvy-humanizer-dv ``` ### Technical Analysis The Quick Start instructions install two third-party ClawHub skills by package name without specifying immutable versions, cryptographic digests, verified source locations, or integrity checks. Neither dependency's implementation is included in this repository, so its effective instructions and executable behavior cannot be reviewed from the audited artifact. Because package names resolve to mutable registry content, future releases or compromised package entries could differ from the versions originally intended by the project author. This creates a supply-chain trust boundary between the reviewed documentation and the code or instructions ultimately installed. ### Attack Path 1. An attacker compromises a maintainer account, registry entry, publication pipeline, or another distribution component for either named skill. 2. The attacker publishes a modified package under the expected package name. 3. A user follows the documented Quick Start commands. 4. ClawHub resolves and installs the current mutable package because no immutable version or digest is specified. 5. The malicious package is loaded or invoked with the permissions available to the host agent. 6. Its resulting capabilities depend on the installed package and host configuration, and could include unauthorized file, network, tool, or agent-context access. ### Impact Assessment The immediate project contains no executable implementation, so no direct malicious behavior is established. The potential impact originates from the externally resolved dependencies. If one is compromised, its payload may operate with the invoking agent's permissions and affect content, accessible files, network resources, tool calls, or session behavior ...[truncated 125 chars]
Remediation
## Remediation Suggestions - Pin each dependency to an immutable, reviewed version and preferably a cryptographic digest. - Document the authoritative registry namespace, publisher identity, and expected integrity value. - Verify signatures or checksums before installation and reject mismatches. - Vendor reviewed dependency contents into the project when practical so the complete behavior can be audited. - Use a lock file or equivalent reproducible dependency manifest if supported by ClawHub. - Re-audit dependencies before updating pinned versions. - Run installed skills with least privilege, restricting filesystem, network, credential, memory, and tool access to the minimum required. - Add installation guidance warning users not to accept unexpected publisher changes or unresolved package substitutions.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Static analysis

No suspicious patterns detected.