ClawHub
Back to skill

Security audit

Defluff

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only email triage skill that analyzes provided emails and shows no hidden code, persistence, or data exfiltration behavior.

Install only if you are comfortable letting the configured OpenClaw model inspect the email content you provide. Treat AI-authorship labels, guessed prompts, and scam classifications as helpful triage signals rather than proof.

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
  • Rogue AgentSelf-Modification, Session Persistence
  • 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 (5)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest says the skill detects AI authorship, guesses likely LLM prompts, classifies urgency, extracts intent, identifies scam patterns, mirrors the email's language, and handles single/thread/batch inputs. The README instead states the skill extracts actual intent into 3–5 bullets and emphasizes a fixed prompt, omitting and thereby narrowing the claimed behavior relative to the manifest.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 13)May include surrounding context.

bash
# copy into your local skills directory
mkdir -p ~/.openclaw/workspace/skills/defluff
cp SKILL.md ~/.openclaw/workspace/skills/defluff/

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The trigger conditions are broad enough to match generic summarization or triage requests whenever a user pastes text, increasing the chance this skill is invoked outside a clear email-analysis context. That can cause misclassification of non-email content, unexpected structured output, or accidental application of scam/authorship judgments where the user only wanted summarization.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Forcing output to mirror the email's language without user choice can violate user expectations, enterprise locale requirements, or safety workflows that depend on a preferred language. In multilingual or monitoring contexts, this may reduce operator comprehension and hide important scam indicators from the actual user if they do not read the source language well.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The manifest explicitly states that output mirrors the email's language, but the README defines fixed English output labels and categories such as 'Actions', 'Triage', 'Act now', 'Reply needed', 'FYI', and 'Noise'. That suggests standardized English output rather than language mirroring.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.