Back to skill

Security audit

freqtrade-tools

Security checks for vulnerabilities and agentic risk

Overview

These Freqtrade shortcuts fit their stated purpose and require confirmation for disruptive actions, though their live-trading safety documentation needs correction.

Use a dedicated dry-run project and verify that FT_PROJECT and FT_CONFIG identify the same bot. Controls affect the entire project. Set FT_CONFIRM only for an action you have approved, and remember that --erase deletes cached market data. The documented live-mode override does not work. Your deployed configuration was not included in this review.

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
  • 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
Findings (4)

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/bash-zsh-aliases.md (reported line 203)May include surrounding context.

echo "Open http://localhost:8080"; }

text

`ftlogs`, `ftstatus`, `ftlist`, and `ftui` are read-only and need no
confirmation. `ftstop`, `ftstart`, and `ftrestart` do.

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/windows-equivalents.md (reported line 221)May include surrounding context.

function ftui { Start-Process "http://localhost:8080" }

text

`ftlogs`, `ftstatus`, `ftlist`, and `ftui` are read-only and need no
confirmation. `ftstop`, `ftstart`, and `ftrestart` do.

---

Bot-control documentation promises a live-trading confirmation override that the guard never permits.

Low
Category
Excessive Agency
Confidence
100% confidence
Finding

L186–L187 state that setting FT_CONFIRM=yes allows bot-control commands against a live bot, and L55 gives the same instruction. However, L53–L56 unconditionally return failure when dry_run is false; FT_CONFIRM is checked only afterward for destructive actions. The implementation therefore contradicts the documented override, although its unconditional refusal is consistent with the manifest's live-bot restriction.

Content

No source excerpt is available for this finding.

Documentation promises a live-bot confirmation override, but the guard unconditionally rejects live configurations.

Low
Category
Excessive Agency
Confidence
100% confidence
Finding

L017–L019, L045–L046, and L198–L199 state that explicit confirmation permits control of a live bot, and L060 instructs users to set FT_CONFIRM=yes to proceed. However, L058–L061 return false whenever dry_run is false, regardless of FT_CONFIRM; the confirmation check at L064 only applies after live configurations have already been rejected. The implementation follows the manifest's stricter refusal policy, but directly contradicts the file's safety documentation.

Content

No source excerpt is available for this finding.