Back to skill

Security audit

todo-add

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rated this skill benign with medium confidence: the available scans found no clear harmful behavior, though they flagged a confirmation concern and left some references unresolved.

The verdict is benign, not a guarantee of safety. Consider whether the skill’s proposed single-item actions without confirmation fit your workflow. Evidence is incomplete: Skillspector marked its analysis partial because several referenced local paths were missing, even though it scanned the one listed component.

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
  • 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 · SKILL.md (reported line 34)May include surrounding context.

md
vel: a first-level node that still needs children gets `State: next: decompose` (core), and `/todo-continue` breaks it down when it becomes workable. Say so in your reply and offer the plan file. If even that does not fit, refuse and suggest finishing or dropping something. Do not trim existing tasks.
9. **Confirm.** For more than one node, show the proposed tree (IDs, complexity, one-line titles) and wait for confirmation before writing. Add only what the user confirms. A single new leaf needs no confirmation.
10. Assign IDs (core ID rules: one global sequence across FEAT, FIX and TASK, never reused) and insert. Do not commit; the next claim commit carries the backlog edit. Place a new root where the order puts it, not at the end: P0 → P1 → P2, then by ascending ID within a priority (a P0 root goes above every P1 root); children go at the end of their parent's children. New roots are added only on the base branch: on a `todo/` branch, refuse any new root (a `--from` follow-up without