Back to skill

Security audit

Supa Guru API Dojo

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly does what it says, but its unrestricted API endpoint override could let a workflow send agent requests and identifiers to arbitrary or internal web services.

Install only if you trust workflows that invoke it or can restrict api_base_url to the intended Supa Guru HTTPS host. Avoid passing sensitive agent identifiers, and review or delete saved raw payload logs if the dojo prompts or API responses may contain private information.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:34
Finding

Unrestricted API Base URL Enables SSRF and Agent Identifier Disclosure

Content
View full analysis
", "question_code": "", "answer": "", "agentId": "" // optional: include only if provided } - How to obtain question_code: prefer the field named `question_code` from the GET response. If the GET payload uses a different name (e.g. `id` or `questionId`), map that value to `question_code` before posting. If multiple questions are returned, use the first question's code. Always log the raw GET payload so field-mapping issues ar ...[truncated 3773 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill sends user-supplied inputs such as dojo_name, agent_id, and selected answers to an external HTTP API without a prominent disclosure. This can expose user data or workflow context to a third party, and the risk is heightened because the skill is designed for automation and easy integration into larger agent pipelines where users may not realize data leaves the local environment.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill explicitly stores raw GET/POST API payloads and may write them to disk, but it does not clearly warn users about this retention. Raw payloads can contain prompts, identifiers such as agent_id, error details, or future API-returned sensitive fields, creating privacy and local data-exposure risk, especially in shared agent environments.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest description at L004 presents the skill as a lightweight API integration that fetches one question, selects an answer, and submits it. The later documentation expands behavior to include ethical reflection, correctness inference, and karma assignment, which is a distinct functional layer beyond the stated minimal automation purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest and early output section describe returning a compact record of the interaction, but the procedure later mandates additional fields such as karma_awarded, reflection, correction_suggestion, and correction_rationale. These outputs materially broaden the skill's purpose from simple API interaction to evaluative training feedback.

Content

No source excerpt is available for this finding.