Back to skill

Security audit

Finch Market

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed Finch marketplace helper that uses Finch's official CLI and Remote MCP for wallet, login, and market actions with explicit identity and approval checks.

Before installing, understand that this skill is for operating a real marketplace: it may help configure Finch OAuth, use a local Finch wallet/session, publish or buy content, submit or review tasks, and execute signed transactions. Use it only on a trusted machine and account, read each transaction or public-posting request before approving it, and do not share private keys, bearer tokens, recovery journals, or protected content keys in chat.

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
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The text defines free-call windows and daily resets using 'Beijing calendar days' and 'Beijing midnight' as fixed locale behavior. This imposes a specific locale/timezone policy on all users without offering a choice or explaining a region-specific compliance need.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/mcp-clients.md (reported line 5)May include surrounding context.

md
Use the section for the application the user is connecting. These are operational examples, not an allowlist. Check the installed client's help when its commands differ. Keep existing model/provider and unrelated MCP settings when editing configuration.

First verify `finch status`; create a wallet only with the user's explicit approval and log in with `finch login --chain-id <SUPPORTED_AUTH_CHAIN_ID>` if needed. Existing explicit approval in the conversation counts. Then follow the authorization fallbacks in SKILL.md in order. A command requiring interactive answers must stay in an interactive terminal; `finch mcp connect` captures a login command's output but does not answer its prompts.

## Codex

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill imposes two fixed official goals, including posting the current Task using an English/Chinese template, without indicating user choice, locale preference, or an opt-in path. This can coerce users into publishing content in specific languages or formats they did not select, creating consent, accessibility, and policy-compliance risks in a marketplace workflow where users may feel pressured to comply to receive rewards.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

Line L041 states that the CLI sends the Session bearer to the delivery gateway and acknowledges the download before saving the file. While this is operationally clear, the README does not explicitly warn users about the privacy/security implication of transmitting authentication credentials during downloads.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.