Back to skill

Security audit

Book Of The Day

Security checks for vulnerabilities and agentic risk

Overview

The skill itself is a simple daily book recommendation, but its install notes tell users to run an unspecified publisher-provided installer outside the reviewed package.

Prefer the reviewed `clawhub install book-of-the-day` path and manually review any cron/channel settings before enabling daily pushes. Do not run a separate publisher-supplied one-command installer unless its full contents, source, version, and checksum/signature are available and trusted.

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
INSTALL.md:28
Finding
Execution of an Unreviewed Publisher-Provided Installation Command## Vulnerability Details **File Location**: `INSTALL.md`, lines 28–30 **Vulnerability Type**: Unauthenticated and unpinned publisher-provided installation command **Risk Level**: Medium **Relevant excerpt translated into English from the source documentation:** ```text If the publisher has not made the default API available, the publisher will provide a “one-command installer” or an API address. - One-command installer: Copy and execute it directly. It will automatically complete the installation and write the configuration. ``` ### Technical Analysis The documentation instructs users to copy and execute an installation command supplied separately by the publisher. The command is not included in the audited package, pinned to a specific immutable version, protected by a published checksum, or otherwise made available for inspection as part of this artifact. Consequently, the effective installation payload can differ from the content reviewed during this audit. A compromised publisher account, delivery channel, website, or other command-distribution mechanism could substitute a malicious command. This creates an unsafe software supply-chain boundary because users are explicitly encouraged to execute content whose provenance and behavior cannot be verified from the package. ### Attack Path 1. An attacker compromises or impersonates the publisher or the channel used to distribute the one-command installer. 2. The attacker supplies a command that downloads or invokes an attacker-controlled payload. 3. A user follows the instructions in `INSTALL.md` and executes the command without being able to compare it against the audited artifact. 4. The malicious command runs with the privileges of the user executing the installer. 5. The payload may access user-readable data, modify OpenClaw configuration, steal locally available credentials, install additional components, or establish persistence where the user's permissions allo ...[truncated 510 chars]
Remediation
## Remediation Suggestions - Remove the instruction to execute an unspecified command supplied outside the audited package. - Include the complete installer in the published project so users and reviewers can inspect it before execution. - Pin all downloaded artifacts and dependencies to immutable versions or content-addressed identifiers. - Publish and verify cryptographic checksums or digital signatures before running any downloaded installer. - Distribute installation assets only through an authenticated official source with HTTPS and appropriate release-signing controls. - Avoid shell pipelines or commands that download content and execute it immediately. - Separate download, verification, and execution into explicit steps so users can inspect the payload. - Ensure the installer operates with least privilege and clearly documents every file and configuration entry it modifies.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (1)

Natural-Language Policy Violations

Medium
Confidence
95% confidence
Finding
The scheduled message is fixed as "给我今日之书", and the document title and usage examples are entirely in Chinese. Under the policy for natural-language constraints, forcing a specific language without user opt-in is a violation unless the locale restriction is explicitly justified or alternatives are provided.

Static analysis

No suspicious patterns detected.