Back to skill

Security audit

openclaw-nextcloud

Security checks for vulnerabilities and agentic risk

Overview

This Nextcloud integration has broad account powers, but its sensitive operations are disclosed and require user-directed approval.

Use a dedicated, revocable app password and understand that it exposes the whole account. Confirm exact targets before changes; prefer read-only, password-protected, expiring shares. Uploads and comment deletion depend on the agent honoring confirmation instructions. Keep HTTP limited to isolated development. This review investigated material scanner gaps statically; it did not test a live account or exhaustively audit every bundled dependency.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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)

T09 · Insecure Skill Coding Practices

Warning
Location
.
Finding

Destructive upload and comment deletion bypass runtime confirmation

Content
View full analysis

Impact

An invocation without the required user approval can overwrite remote content or delete a comment using the account-level app password.

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (16)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The supplied command is a skill-information lookup, not an implementation of the declared Nextcloud management functionality. No Nextcloud API calls or resource-management operations are shown. This assessment is limited to the supplied code chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Most visible functionality matches the declared purpose. However, Shares.createLink and the shares create-link command create public shares using shareType=3, with permission values of 1 (read) or 15 (read/create/update/delete). Password protection and expiration are optional. This changes who can access or modify files and is materially distinct from the described upload/download functionality. Share listing and deletion are also implemented but undeclared. Contact access through CardDAV and Deck labels/comments are supporting parts of the stated resource-management purpose, not separate mismatches.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

Most visible operations match the declared Nextcloud management purpose. However, the shares create-link branch invokes Shares.createLink with a file path, permissions, optional password, and expiration date. Creating sharing links changes who can access files and is a security-relevant capability beyond the described uploading/downloading of files. Listing and deleting shares are also not disclosed. The supplied chunk does not show the underlying API implementations, so this assessment is limited to the operations it dispatches.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

Public-link creation and share administration are distinct access-control capabilities not explicitly disclosed by the description's file-management and upload/download functionality. Creating a public share can expose content to others, making this more than a supporting implementation detail. The chunk contains command examples rather than the CLI implementation, so this assessment concerns the capabilities those examples invoke.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
89% confidence
Finding

Most behavior matches the declared purpose. Credential access, authenticated network requests, local upload reads, and download writes are supporting implementation details rather than mismatches. However, Shares.createLink creates public shares (shareType 3), optionally granting create, read, update, and delete permissions. This changes access to stored files and is materially distinct from the explicitly described upload/download functionality. The supplied chunk implements these share operations and lists their CLI flags, although it ends before the corresponding main-dispatch code can be verified.

Content

No source excerpt is available for this finding.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 44)May include surrounding context.

md
- **Node.js 24+** on PATH (`node scripts/nextcloud.js` in Claude Code or a

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 145)May include surrounding context.

md
node scripts/nextcloud.js <command> <subcommand> [options]

Code capability 'network' detected in index.js but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
78% confidence
Finding

Assuming the reported permission mismatch is accurate, authenticated network requests occur outside the declared capability contract. Network access is central to the advertised functionality, so this is a permission-disclosure issue rather than evidence of malicious exfiltration; nevertheless, reviewers should be able to approve the destination and credential transmission explicitly.

Content

No source excerpt is available for this finding.

Code capability 'env' detected in index.js but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
78% confidence
Finding

Assuming the reported permission mismatch is accurate, the skill accesses account credentials without declaring that capability, leaving its permission contract incomplete. These reads are necessary for an authenticated Nextcloud client and do not themselves demonstrate credential theft, but undeclared secret access weakens informed approval and least-privilege review.

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · scripts/nextcloud.js (reported line 16242)May include surrounding context.

js
her.readOnly();
   * // pass view to callbacks — it stays in sync automatically
   * view.matches(expr);       // ✓
   * view.getCurrentTag();     // ✓
   * // view.push(...)         // ✗ method does not exist — caught by TypeScript
   */
  readOnly() {
    return this._view;
  }
};

// node_modules/is-unsafe/src/contexts/html.js
var HTML_PATTERNS = [
  {
    id: "html-script-open",
    description: "<script opening tag",
    pattern: /<script[\s>/]/i
  },
  {
    id: "html-script-close",
    description: "</script closing tag",
    pattern: /<\/script[\s>]/i
  },
  {
    id: "html-javascript-protocol",
    description: "javascript: URI scheme (with optional whitespace/encoding)",
    // Handles j&#x61;vascript:, j\u0061vascript:, and whitespace variants
    pattern: /j[\t\n\r ]*a[\t\n\r ]*v[\t\n\r ]*a[\t\n\r ]*s[\t\n\r ]*c[\t\n\r ]*r[\t\n\r ]*i[\t\n\r ]*p[\t\n\r ]*t[\t\n\r ]*:/i
  },
  {
    id: "html-vbscript-protocol",
    description: "vbscript: URI scheme",
    pattern:

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 464)May include surrounding context.

md
2. **If user sets a default:**
   - Remember `default_task_calendar` and/or `default_event_calendar`
   - Use automatically for subsequent operations without asking

3. **If user declines to set a default:**
   - Ask again next time they create a task/event without specifying a calendar

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 488)May include surrounding context.

md
2. **If user sets a default:**
   - Remember `default_addressbook`
   - Use automatically for subsequent operations without asking

3. **If user declines to set a default:**
   - Ask again next time they create a contact without specifying an address book

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 511)May include surrounding context.

md
2. **If user sets a default:**
   - Remember `default_deck_board` (store the board id and title)
   - Use automatically for subsequent operations without asking

3. **If user declines to set a default:**
   - Ask again next time they work with cards without specifying a board

The documented HTTPS-by-default guarantee contradicts the unconditional localhost HTTP exception.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

L0012-L0014 state that HTTPS is required by default and describe OPENCLAW_ALLOW_HTTP=1 as enabling local plaintext development. However, L0092-L0093 exempt localhost and loopback addresses without that opt-in, allowing the Basic Auth credentials to be transmitted over HTTP under the default configuration. This contradicts a documented credential-transport safeguard.

Content

No source excerpt is available for this finding.

The skill can publish public file shares and grant anonymous write/delete access, beyond the manifest's described file operations.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The skill can create public bearer-link shares, including permissions value 15, which requests anonymous read, create, update, and delete rights where the server permits them. This sensitive capability is not disclosed in the supplied description, although the exact-action confirmation requirement and read-only default substantially reduce accidental exposure.

Content

No source excerpt is available for this finding.

The skill can publish public file links with write/delete access, beyond the described file upload/download operations.

Medium
Category
Excessive Agency
Confidence
87% confidence
Finding

The manifest describes managing resources in the user's Nextcloud instance, including uploading and downloading files, but does not disclose public sharing or changing third-party access. At L19143–L19168, createLink creates a public share (shareType 3), optionally grants create/update/delete permissions, and allows omission of both password and expiration; this exposes resources to link holders rather than merely managing their contents. The capability is reachable through the shares create-link command at L20179–L20194, although explicit confirmation is required.

Content

No source excerpt is available for this finding.