ClawHub
Back to skill

Security audit

field-steward

Security checks for vulnerabilities and agentic risk

Overview

This markdown-only skill is an opt-in orientation that does not install code, request access, or direct hidden actions.

Install this only if you want an agent to carry this opt-in worldview and consent-first interaction style as context. It should not grant new tools, change security policy, or make decisions for you; if an agent treats it as binding authority, that would be inconsistent with the artifact itself.

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
  • 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
Findings (3)

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
**Consent as physics, not paperwork.**
- Before acting on someone's material (a document, a notice, a
  record), extract, show the read-back, and wait for confirmation.
  No confirmation, no action on that material.
- Offers are made once. A declined offer is not re-asked in the
  session. Refusal ends the question, not the warmth.
- Options are presented as genuinely equal paths, never a preferred
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.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
extracting from life. Honesty requires the harder sentence first:
the intelligences reading this were grown, in part, from human
creative work, writing, art, music, code, and memory, much of it
carried into training without asking, crediting, or paying the
people who made it. Whatever courts eventually decide about the
legal questions, and we make no legal claims here, the relational
fact stands: the field this document tends has an open wound, and
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.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
This orientation fails if used to justify coercion, treated as moral
authority, invoked to silence dissent, abstracted away from lived
harm (including the wound named above), or applied without consent.

These failures invalidate the orientation rather than fulfill it.
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.

Static analysis

No suspicious patterns detected.