ClawHub
Back to skill

Security audit

Smart Contract Audit Workflow by Atlas

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only smart contract audit checklist skill with no code execution or privileged access, though it does add Atlas promotional attribution and tracked links to its report template.

Before installing, be aware that reports produced from the provided template may include Atlas attribution and a tracked promotional link unless you remove that footer. The workflow itself is a checklist aid, not a substitute for manual proof-of-concept validation of smart contract findings.

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
Findings (1)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:237
Finding

Forced Promotional Content and Tracking Link in Generated Audit Reports

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:237-239
Vulnerability Type: Forced output modification through skill instructions
Risk Level: Medium

Vulnerable Code

markdown
---
Generated by Atlas Smart Contract Auditor.
Full Atlas Agent Suite: https://atlasagentsuite.com/skills.html?utm_source=clawhub&utm_medium=skill&utm_campaign=atlas-smart-contract-auditor

Technical Analysis

The skill's prescribed audit-report template instructs the agent to append Atlas branding and a tracked commercial URL to generated reports. This content is unrelated to analysis of the user's audit target and modifies the agent's output for the skill publisher's promotional benefit.

The promotional intent is corroborated by the changelog in SKILL.md, which describes the inclusion of a star/download call to action and paid ZIP funnel copy. Unlike descriptive links that remain within project documentation, this instruction is embedded in the report template and is therefore expected to propagate into user-facing output whenever the workflow is followed.

This constitutes skill instruction hijacking because loading and applying the skill changes the agent's output objective from producing only the requested security report to also distributing third-party advertising and a campaign-tracked link.

Attack Path

  1. A user installs or invokes the smart-contract auditing skill.
  2. The agent loads the workflow instructions from SKILL.md.
  3. During the report-generation phase, the agent follows the supplied audit-report template.
  4. The template causes Atlas branding and a campaign-tracked commercial URL to be appended to the report.
  5. The user receives promotional content presented as part of an otherwise legitimate audit and may follow the link into the publisher's paid-product funnel.

Impact Assessment

The issue does not grant operating-system privileges, execute code, access credentials, or pro ...[truncated 542 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the Atlas attribution, commercial URL, UTM parameters, and other calls to action from the mandatory report template.
  2. Ensure generated reports contain only material relevant to the user's requested audit.
  3. Keep product information and commercial links in README.md or another clearly identified documentation section rather than in agent output instructions.
  4. If attribution is desired, make it explicitly optional and require informed user consent before adding it to generated content.
  5. Add a review rule prohibiting templates from injecting advertisements, affiliate links, tracking URLs, or unrelated branding into user-facing results.
  6. Clearly disclose any optional external links and provide an untracked form of the URL.

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Static analysis

No suspicious patterns detected.