ClawHub
Back to skill

Security audit

wilma

Security checks for vulnerabilities and agentic risk

Overview

The skill’s access to school records and saved logins fits its stated purpose, and both scanner alerts misread protective instructions.

Connect only accounts you are authorized to access and protect the computer or secret settings holding passwords and optional MFA keys. Requests cover all connected children unless narrowed. School information read by your assistant reaches its AI provider, as the [publisher explains](https://github.com/aikarjal/wilmai#credentials--privacy). Logout keeps saved credentials; use account removal or config clearing to remove them. Installation and updates execute externally distributed npm code.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • 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 (2)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 75)May include surrounding context.

md
es).
  2. Ask the user to add `WILMA_USERNAME` and `WILMA_PASSWORD` to your secret or environment settings, plus `WILMA_TOTP_SECRET` if the account uses two-step verification, and set `WILMA_TENANT` to the Wilma address they picked. Every command then works without a saved login. Alternatively: `wilma login --tenant <url> --username <name> --password-stdin`.
- Don't run `wilma` without arguments to look things up: in a terminal it opens a menu for people (from a program it prints the help).
- **Never ask the user to type their Wilma password into the chat.**
- **Children on different Wilmas** (e.g. a city school and a private school): the user adds each login on the login page ("Add another Wilma") or with another `wilma login`. All saved logins are used together; children then carry a `wilma` field naming their Wilma. `wilma accounts` lists the logins.

Commands continue the last Wilma session instead of logging in each time (each login would log the parent out of Wilma in their own b

Unsafe Defaults

Medium
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.

Content

Scanner excerpt · SKILL.md (reported line 198)May include surrounding context.

md
new>` on stderr (at most once a day). Finish the user's request first, then run `wilma update`, or tell the user if you can't install software here.
- The 1.x spellings (`kids list`, `tenants`, `attendance list`, `<command> list`, `messages read <id>`, `news read <id>`, `news resource download …`, `--all-students`) still work.
- **TLS errors on managed machines.** If a command fails with code `network` and a `cause` such as `UNABLE_TO_GET_ISSUER_CERT_LOCALLY`, the network is intercepting TLS and re-signing certificates with a private root CA that Node does not trust. Read the `hint` and report it rather than retrying. The fix is to run the CLI with `NODE_USE_SYSTEM_CA=1` (Node >=22.19/>=24.6), or `node --use-system-ca "$(command -v wilma)"` (Node >=22.15). Never suggest `NODE_TLS_REJECT_UNAUTHORIZED=0`; it disables verification entirely. A working `npm install` does not prove TLS is healthy — the npm registry is commonly exempt from inspection.

## Actionability guidance (for parents)

Static analysis

No suspicious patterns detected.