Back to skill

Security audit

Skill

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed local companion integration, with ordinary setup risks from installing and authorizing a separate daemon.

Before installing, verify you trust the drakeling package source, consider pinning a known version, and configure DRAKELING_API_TOKEN only for this skill if your OpenClaw setup supports it. The skill should only be used with the local Drakeling daemon and should not be granted unrelated secrets.

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:37
Finding
Unpinned Third-Party Package Installation## Vulnerability Details **File Location**: `SKILL.md:37`; `references/setup.md:9-13` **Vulnerability Type**: Unpinned third-party dependency installation **Risk Level**: Medium ### Vulnerable Code `SKILL.md:37`: ```bash pipx install drakeling ``` `references/setup.md:9-13`: ```bash pipx install drakeling # recommended — isolated environment pip install drakeling # standard pip uv tool install drakeling # uv ``` ### Technical Analysis The installation instructions retrieve the currently published `drakeling` package without specifying an exact version, verifying a cryptographic hash, or using a reviewed lockfile. Consequently, the installed code may differ from the version that was available when this Skill was audited. Installation through `pip`, `pipx`, or `uv` can execute package-controlled build or installation behavior. The resulting package also supplies the `drakelingd` executable that users are instructed to run. According to the supplied documentation, that daemon participates in LLM credential setup, creates identity material and an API token, and operates with the invoking user's privileges. A compromised package release, package registry account, or dependency could therefore introduce arbitrary code outside the reviewed Skill files. The external package implementation is not included in this project, so this audit does not establish that the current package is malicious. The vulnerability is the absence of dependency pinning and integrity verification in the installation process. ### Attack Path 1. An attacker compromises the package publisher, distribution account, package registry, or a transitive dependency and publishes a malicious release. 2. A user follows the documented unpinned installation command. 3. The package manager resolves and installs the attacker-controlled release rather than a previously audited version. 4. Malicious code executes during package install ...[truncated 1109 chars]
Remediation
## Remediation Suggestions 1. Pin installation instructions to a specific, reviewed release, for example: ```bash pipx install "drakeling==1.0.6" ``` 2. Publish and verify cryptographic hashes for the approved distribution artifacts. Where supported, use a requirements file with exact versions and `--require-hashes`. 3. Lock all transitive dependencies and review lockfile changes before updating the approved release. 4. Document the authoritative package registry, publisher identity, and source repository so users can detect dependency confusion or typosquatting. 5. Prefer reproducible, signed release artifacts and document signature verification. 6. Test new versions in an isolated environment before changing the documented pin. 7. Run the daemon as an unprivileged user with access limited to the files and credentials required for its legitimate function. 8. Avoid exposing unrelated secrets to the daemon process and keep the Drakeling token in skill-scoped configuration rather than a global environment block whenever possible.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (3)

Vague Triggers

Medium
Confidence
88% confidence
Finding
The trigger description is broad enough that ordinary mentions of a 'companion creature' or generic requests to 'check in' could invoke this skill unexpectedly. Because the skill has outbound network permission and sends authenticated requests to a local daemon, accidental activation could cause unintended state-changing actions such as sending care or disclosing creature status without clear user intent.

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The setup guide shows the skill depends on a separate local daemon, generated identity keys, and a local HTTP API secured by a token, which materially expands the operational and trust boundary beyond the simple creature-care description in the manifest. This mismatch can mislead users and reviewers about what components are actually involved and what local services and secrets the skill relies on, increasing the chance of unsafe deployment or overbroad trust.

Context-Inappropriate Capability

Medium
Confidence
88% confidence
Finding
The documentation instructs users to retrieve a local API token from disk and inject it into the skill environment, giving the skill access to authentication material not obviously justified by the narrow 'check on/send care' functionality. Even if intended for legitimate local authentication, requiring secret handling inside the skill increases the risk of token exposure, misuse, or reviewers underestimating the skill's effective privileges.