Back to skill

Security audit

Reducto

Security checks for vulnerabilities and agentic risk

Overview

The Reducto workflow is coherent, but its fallback setup can execute a changing remote installer without clear approval or limits on local changes.

Review the fallback installation before enabling this skill. Prefer manually installing a verified CLI release and require approval for subsequent local setup changes. Check the CLI's skill-synchronization settings, and use only documents you authorize OOMOL and Reducto to process. Installer contents and live connector behavior were not verified; the inspected artifact does not establish malicious intent.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (5)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied command executes an Oomol installation script and contains no Reducto interaction. Nothing supplied establishes that this installer is a supporting dependency for the declared functionality. The remote script's contents and further effects cannot be determined from this code chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
84% confidence
Finding

Authentication can be a supporting implementation detail, but the supplied chunk consists entirely of a login command and does not establish that it authenticates to Reducto. As supplied, its primary behavior does not match the declared Reducto search/read purpose. The command alone does not establish credential theft or other malicious behavior.

Content

No source excerpt is available for this finding.

External Script Fetching

High
Category
Supply Chain
Confidence
98% confidence
Finding

Piping a remotely fetched, unpinned script directly into bash executes whatever the endpoint serves with the agent process's permissions. Compromise of the installer distribution infrastructure could therefore expose local files and credentials or enable persistent modification; HTTPS does not verify that the script is an approved release. The conditional setup path limits exposure but does not remove this supply-chain risk, and the PowerShell example has the same execution pattern.

Content

Scanner excerpt · SKILL.md (reported line 57)May include surrounding context.

The setup workflow executes downloaded installation scripts without warning about remote-code execution or system changes.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The missing-CLI recovery instructions direct the agent to execute downloaded scripts without explaining the resulting local code execution or obtaining approval for installation. Although installation supports the skill's legitimate purpose and is conditional, a routine connector request can consequently trigger system changes outside document processing.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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 78)May include surrounding context.

md
- **`OOMOL_INSUFFICIENT_CREDIT`** — OOMOL billing stop. Recharge at `https://console.oomol.com/billing/token-recharge` before retrying.

- **HTTP 402 without `OOMOL_INSUFFICIENT_CREDIT`** — the third-party account itself has no active plan or credit. Report the provider error message as is; do not ask the user to recharge OOMOL.

## Resources