ClawHub
Back to skill

Security audit

Candor Agent UI

Security checks for vulnerabilities and agentic risk

Overview

The scanners rate this skill benign with high confidence: all seven listed components were fully inspected, with one low-probability checklist finding that the reviewing analyzer disputed.

The verdict is benign, and Skillspector reports complete inspection of all seven listed components. Consider that the single checklist-related flag was disputed and marked as expected; benign does not mean risk-free.

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 (1)

Autonomous Decision Making

Medium
Category
Excessive Agency
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.

Content

Scanner excerpt · references/review-checklist.md (reported line 64)May include surrounding context.

md
- [ ] Every status is a word beside its colour or dot; nothing is colour alone (status vocabulary)
- [ ] Statuses map to the nine run statuses; failed ones carry an error, blocked ones name what blocks them (status vocabulary)
- [ ] A tool call shows what it did in plain words, its status, and a duration only once measured; its input and output are text (tool call states)
- [ ] Folded turns open; no approval, error or real change is hidden inside a fold (folding turns)
- [ ] While working, the screen shows the current step and what it's waiting on, and the person can stop (while it works)
- [ ] Work finished while away leads with what needs the person, then failures, then results; nothing is auto-dismissed (finishing while away)