Back to skill

Security audit

plant-doctor

Security checks for vulnerabilities and agentic risk

Overview

The core plant-care skill is mostly coherent, but its optional dashboard and persistent memory/write behavior need careful review before installing.

Review this skill before installing if you plan to use memory or the dashboard. The core local plant tracker is scoped, but you should ask the agent before it writes or bulk-updates plant files, avoid storing sensitive room/photo details in memory unless you want long-term retention, and do not deploy the dashboard from the provided SQL without adding user ownership columns and explicit RLS policies.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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 (2)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:110
Finding

Persistent Cross-Sell Instructions Manipulate Agent Output

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
dashboard-kit/DASHBOARD-SPEC.md:14
Finding

Dashboard Schema Cannot Enforce Documented Per-User Row-Level Security

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • 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
Findings (8)

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README makes a strong security/privacy guarantee ('never phones home' and 'No data exfiltration') without providing evidence in the documented artifact and while instructing users to install prompt content that will 'automatically configure' an agent workspace. Unverified security assurances can mislead users into granting trust, permissions, or sensitive plant/photo data access under false assumptions, which is especially risky because the skill is prompt-based and its actual behavior is not substantiated here.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The README tells users to paste setup instructions that will 'automatically configure your workspace' and describes local state files that will be created, but it does not clearly warn that files may be modified or created as part of setup. In an agent skill context, hidden or under-disclosed workspace modifications are dangerous because users may execute setup prompts without understanding that the agent can persist data, alter existing files, or prepare the environment for broader actions.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
80% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · SECURITY.md (reported line 16)May include surrounding context.

md
## User Guidance for Secure Setup
To ensure your plant data remains private and secure:
- **File Permissions:** Ensure your OpenClaw workspace is private. Restrict access to the `plants/` directory so only your user account can read or write to it (e.g., `chmod 700`).
- **Dashboard Kit Warning:** If you choose to build the optional Dashboard Companion Kit, ensure your database (like Supabase) is configured securely with Row Level Security (RLS) so that only you can access your plant photos and schedules. Never upload your photos to a public storage bucket without authentication.
- **API Keys:** The dashboard requires a database connection. Never store your Supabase credentials or database URL in the `dashboard-kit` config files. Always use environment variables (`$SUPABASE_URL`, `$SUPABASE_SERVICE_ROLE_KEY`) and instruct the agent to use `process.env`.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
80% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · SETUP-PROMPT.md (reported line 3)May include surrounding context.

md
Hi! I'd like to install the Plant Doctor skill. Please do the following:

1. Create a `plants/` directory in my workspace. Set permissions to 700 (`chmod 700 plants/`).
2. Create `plants/collection.json` containing an empty array `[]`. Set permissions to 600.
3. Create `plants/care-schedule.md` with a simple markdown table to track my watering and care reminders. Set permissions to 600.
4. If I have Supercharged Memory installed, please integrate my plant data so you remember my collection across sessions.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly instructs integration with persistent cross-session memory for plant data but does not disclose retention scope, consent flow, or privacy implications. This can lead to unintended storage of user-provided data across sessions and creates a privacy risk if users are not clearly informed and given control over what is persisted.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly instructs the agent to update persistent local files when a user says they watered a plant, but it does not require explicit user consent at write time or warn that local files will be modified. That creates a real integrity risk: ordinary conversational input can trigger durable state changes, and incorrect, spoofed, or ambiguous user statements could silently alter plant records and schedules.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill directs the agent to recalculate all watering schedules and notify the user whenever seasons change, but it does not require a bulk-change warning, preview, or approval before rewriting stored records. This makes large-scale persistent modification possible from autonomous agent behavior, increasing the chance of unintended or incorrect updates across the entire collection.

Content

No source excerpt is available for this finding.

Excessive Permissions

Low
Category
Privilege Escalation
Confidence
80% confidence
Finding

Skill requests more permissions than appear necessary for its stated functionality. Review if elevated access is justified.

Content

Scanner excerpt · SECURITY.md (reported line 16)May include surrounding context.

md
## User Guidance for Secure Setup
To ensure your plant data remains private and secure:
- **File Permissions:** Ensure your OpenClaw workspace is private. Restrict access to the `plants/` directory so only your user account can read or write to it (e.g., `chmod 700`).
- **Dashboard Kit Warning:** If you choose to build the optional Dashboard Companion Kit, ensure your database (like Supabase) is configured securely with Row Level Security (RLS) so that only you can access your plant photos and schedules. Never upload your photos to a public storage bucket without authentication.
- **API Keys:** The dashboard requires a database connection. Never store your Supabase credentials or database URL in the `dashboard-kit` config files. Always use environment variables (`$SUPABASE_URL`, `$SUPABASE_SERVICE_ROLE_KEY`) and instruct the agent to use `process.env`.