ClawHub
Back to skill

Security audit

Support Scripts

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only customer support scripting guide with disclosed, purpose-aligned guidance and no hidden execution, persistence, or credential behavior.

Installers should treat this as a drafting aid, not legal or platform-policy authority. Before loading scripts into helpdesk macros or auto-replies, confirm refund limits, safety-claim escalation, marketplace rules, and trigger definitions with the business owner or compliance reviewer.

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
  • 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
Findings (4)

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 40)May include surrounding context.

md
2. **Identify and rank the top inquiry types.** Cluster the tags into clear scenarios (e.g. "Where is my order?", "Item arrived damaged", "Wrong size"). Rank by volume × handle time × escalation rate so you script the highest-leverage scenarios first. Aim for 15-25 scenarios in v1; resist trying to cover every edge case.

3. **Define brand voice and policy rules.** Set the voice (register, warmth, formality, emoji usage) and the non-negotiable guardrails: what agents can promise on refunds, reships, and timelines without approval, and what must escalate. Write these as plain rules so every script inherits the same boundaries. See `references/tone-and-policy-guide.md`.

4. **Draft scripts per scenario, two variants each.** For each scenario write variant A (standard) and variant B (high-empathy / repeat or upset customer). Lead with acknowledgement, give the answer or action, and end with a concrete next step. Keep chat variants short; keep email variants complete. Use `references/scenario-library.md` for seeds.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The example 'never admit fault on injury claims' embeds a potentially coercive or legally sensitive communication policy directly into a reusable template without context, approval boundaries, or instructions to defer to compliance/legal teams. In a customer-support skill, this can normalize withholding appropriate safety escalation or factual acknowledgment, leading agents to mishandle harm reports and increasing legal, regulatory, and trust risk.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This markdown template includes a trigger field described only as 'when an agent reaches for this script — keywords, tags, or situation,' without requiring specific invocation criteria or exclusion conditions. That can lead authors to define overly broad or ambiguous triggers that overlap with routine support activity.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The file describes each scenario as having a 'trigger' but does not define how these triggers are used, whether they are exact phrases, intent labels, or internal routing conditions. Without explicit scope, constraints, or negative examples, many entries could overlap with ordinary customer language and cause unintended invocation in a skill that relies on these descriptions.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.