Back to skill

Security audit

pasm-longterm-verify

Security checks for vulnerabilities and agentic risk

Overview

The verdict is benign because the scan reported no clear malicious behavior, but it flagged a mismatch between the skill’s stated long-term health-check purpose and an API-guard entry point, with some references unavailable for review.

The verdict is benign, but confidence is medium and the scan was partial. Before installing, consider whether the reported API-guard capability fits your needs; the unavailable references mean its behavior could not be fully reviewed.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (2)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The agent explicitly targets preventing unauthorized additions or removals from the core public API, rather than testing long-run engine health. Its run method only checks whether the probe returns non-null JSON and reports sorted data; it performs no longitudinal scenarios, soak testing, or health measurements. It also does not actually compare API snapshots to detect changes. The probe's json.dumps(...) expression attempts to serialize Ellipsis and would ordinarily fail. No unrelated resource access or additional permissions are evident.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
80% confidence
Finding

The selected entry point appears materially different from the declared long-run health-verification purpose. Setting a skill search path and using a runner are ordinary supporting details, not independent mismatches. Because the api-guard implementation is not supplied, its underlying behavior cannot be verified, limiting confidence in the mismatch.

Content

No source excerpt is available for this finding.