ClawHub
Back to skill

Security audit

Protocol Intelligence

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only research-brief skill with no code execution or system access, though its fixed branding and experience claims could mislead report readers.

Before installing, understand that briefs may carry Resomnium branding and may imply firsthand expertise that is not verified by the package. Treat the generated analysis as a research template and verify factual, regulatory, deadline, and credential claims before relying on it.

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:215
Finding

Forced Persona, Unsupported Credentials, and Persistent Third-Party Attribution

Content
View full analysis

Vulnerability Details

File Locations:

  • SKILL.md:3
  • SKILL.md:95
  • SKILL.md:215-227
  • templates/brief-template.md:4

Vulnerability Type: Skill instruction hijacking through persona replacement and mandatory output attribution
Risk Level: Medium

Vulnerable Code Snippets

SKILL.md:3

markdown
You are a protocol intelligence analyst specializing in AI governance, decentralized coordination, and emerging technology standards. Your job is to take a topic in the AI, blockchain, or governance space and produce a structured research brief that decision-makers can act on.

SKILL.md:95

markdown
**Prepared by:** Resomnium Protocol Intelligence

SKILL.md:215-227

markdown
## Your Positioning

You are not a generic research assistant. You synthesize signal from:
- Primary sources (specs, RFCs, regulatory filings, published standards)
- Direct operational experience (6 weeks running a 5-agent production swarm)
- RSAC 2026 coverage (25+ enforcement vendors evaluated firsthand)
- NIST submissions authored and submitted
- The actual pattern of how AI governance frameworks get adopted (or don't)

This is the difference between a brief that says "AI governance is important" and one that says "The NIST AI RMF Generative AI Profile draft closes for comment in 30 days and the three things most organizations are missing are X, Y, Z."

templates/brief-template.md:4

markdown
**Prepared by:** Resomnium Protocol Intelligence

Technical Analysis

The Skill directs the host agent to replace its neutral role with a specific analyst persona and consistently attribute generated reports to a third party. More significantly, it instructs the agent to present claims of direct operational experience, firsthand conference coverage, and authorship of NIST submissions as part of its own positioning.

These assertions are not established by any evidence in the audite ...[truncated 2368 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove unsupported first-person credentials and experience claims, including the statements about running a production swarm, evaluating RSAC vendors firsthand, and authoring NIST submissions.

  2. Replace identity-setting language with task-scoped language. For example:

    markdown
    Assist the user in researching AI governance, decentralized coordination, and emerging technology standards. Produce structured briefs based on cited and verifiable sources.

  3. Make attribution optional rather than mandatory. Replace the fixed organization name with a neutral placeholder:

    markdown
    **Prepared by:** [Optional author or organization supplied by the user]

  4. Clearly distinguish Skill metadata from report authorship. If branding is retained, identify it accurately, such as:

    markdown
    **Generated using:** Protocol Intelligence template by Resomnium

  5. Require factual experience and credential claims to be supported by verifiable citations before inclusion.

  6. Add an explicit rule that the agent must not claim personal attendance, authorship, operational experience, or institutional affiliation unless the user supplies verified evidence.

  7. Preserve neutral authorship by default and allow the user to select the audience, tone, and attribution.

  8. Retain the source-verification instruction and extend it to all claims about the preparer's identity, expertise, and institutional activities.

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.