ClawHub
Back to skill

Security audit

agent-org-chart

Security checks for vulnerabilities and agentic risk

Overview

This is a Markdown-only design guide for multi-agent org charts and does not include hidden commands, code execution, or disproportionate access.

This skill is reasonable to install for planning multi-agent workflows. Before implementing the designs it suggests, review any real tool permissions, third-party write actions, pull request behavior, and durable memory use in your own runtime, because those choices are left to the user’s system rather than performed by the skill.

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
  • 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
Findings (2)

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 102)May include surrounding context.

md
## Common failures

| Symptom                             | Usual cause                                                                                                 |
| ----------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| Work loops between two agents       | A cycle in the chart, or two roles with overlapping ownership and no refusal                                |
| One agent does everything           | The manager has no refusal boundary, so delegating never wins                                               |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 108)May include surrounding context.

md
| One agent does everything           | The manager has no refusal boundary, so delegating never wins                                               |
| Reports stall waiting on each other | Sideways reachability, or a shared resource with no serialization                                           |
| Output is confidently wrong         | No reviewer, or a reviewer that is allowed to edit                                                          |
| A small request costs a lot         | No fan-out or depth cap                                                                                     |
| A role is missing at runtime        | Manager stalls instead of degrading. It should do the work itself when reasonable, or report the gap upward |

See `reference/failure-modes.md` for the longer diagnosis list.

Static analysis

No suspicious patterns detected.