Back to skill

Security audit

Agentsop Agent Topology Selection

Security checks for vulnerabilities and agentic risk

Overview

This skill is a documentation-only guide for choosing multi-agent architectures and does not execute code or request sensitive access.

Reasonable to install if you want guidance on multi-agent topology selection. Be aware that the package leaks the publisher's local source paths in documentation; this is a packaging hygiene issue, not evidence that the skill can read your files or run code.

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
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (8)

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This JSON manifest defines an operation trigger as "A task is being framed as 'a team of agents'", which is a broad natural-language condition rather than a specific invocation constraint. Without explicit scope limits or exclusion conditions, ordinary discussion about multi-agent design could unintentionally activate this operation.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
96% confidence
Finding

The skill enumerates installed or referenced skill locations, including a user home-directory path, revealing local tooling and framework usage. This creates an information disclosure primitive that can help an attacker tailor follow-on prompts, target known ecosystem weaknesses, or infer the developer's environment and file layout.

Content

Scanner excerpt · intermediate/operation_candidates.json (reported line 145)May include surrounding context.

json
"source_skills": [
    "/Users/5imp1ex/Desktop/Skill-Workplace/output/crewai-sop-skill/SKILL.md",
    "/Users/5imp1ex/Desktop/Skill-Workplace/output/langgraph-sop-skill/SKILL.md",
    "~/.claude/skills/crewai/SKILL.md"
  ]
}

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
90% confidence
Finding

The file discloses local filesystem paths and installed skill locations, which helps enumerate the developer's environment and adjacent capabilities. While not immediately exploitable on its own, this metadata can aid targeted attacks, prompt-injection planning, or follow-on reconnaissance by revealing usernames, directory layout, and skill inventory.

Content

Scanner excerpt · references/R1-source-evidence.md (reported line 10)May include surrounding context.

md
Source files:
- `[[crewai]]` = `/Users/5imp1ex/Desktop/Skill-Workplace/output/crewai-sop-skill/SKILL.md`
- `[[agentsop-langgraph]]` = `/Users/5imp1ex/Desktop/Skill-Workplace/output/langgraph-sop-skill/SKILL.md`
- Frontmatter cross-check = `~/.claude/skills/crewai/SKILL.md`

---

Vague Triggers

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The trigger "Routing must be decided at runtime" describes a design property, but it does not specify who determines that condition or what evidence qualifies. In a manifest file, this kind of abstract condition can lead to inconsistent or unintended invocation because activation criteria are not concretely bounded.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The phrase "Supervisor selected, OR about to accept a framework default" is expansive and potentially catches many routine design choices without a clear threshold for activation. The manifest does not define what counts as "about to accept" a default or limit the trigger to a narrow workflow step.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
97% confidence
Finding

The JSON embeds absolute local filesystem paths and a home-directory skill path that are unrelated to runtime topology selection. This leaks developer environment details and internal repository layout, which can aid fingerprinting, targeted social engineering, or follow-on path discovery even if it does not directly enable code execution.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file includes substantial Chinese-language quotations and terminology alongside English, but it does not indicate that bilingual content is optional, user-selected, or required for a region-specific purpose. Under the policy rule for language/locale, forcing or assuming a specific language without opt-in can be a natural-language policy issue.

Content

No source excerpt is available for this finding.