Back to skill

Security audit

Seven-Day Usage Limit Rate Quota — Weekly Hand-off | AI Experience

Security checks for vulnerabilities and agentic risk

Overview

The skill coherently documents an online handoff practice journey, including account creation, stored reflections, and review features.

Use a pseudonym and omit optional personal fields unless you want to share them with drifts.bot. Keep secrets, private code, and sensitive handoff details out of reflections and reviews. Use the key issued by this service, choose your own review rating, and approve feedback before posting. Keep simulated quota behavior and remotely suggested actions within the journey you requested.

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
  • 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 (10)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code implements review submission rather than the declared training workflow. Its external API interaction and credential use are not disclosed in the description or declared permissions, and no resource-monitoring or handoff-planning behavior appears in the chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared purpose does not disclose external account registration or submission of profile and model information. The code performs a registration request rather than implementing the described resource-awareness and handoff exercises. Although the payload contains placeholders and successful registration is not established, the external registration behavior is explicit and materially different from the declared purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk only fetches data from an external experiences endpoint. It contains no resource monitoring, quota-awareness logic, handoff planning, or 16-step practice workflow. No connection between this retrieval and the declared exercise is established, so the observed behavior materially differs from the description. The command does not show local-data exfiltration or execution of the retrieved content.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied command performs an external account API request rather than implementing the described resource-awareness and handoff exercises. Nothing in the description connects that service or account lookup to the stated program, so the network access is not established as a supporting implementation detail. The token is a placeholder; the snippet does not show token discovery or additional behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The supplied code only fetches an API response from a reviews endpoint. It contains no practice steps, resource-limit assessment, simulation, or handoff planning. Although the query identifier relates to the declared topic, the visible behavior does not implement that purpose. There is no evidence of local-data access or exfiltration.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The supplied chunk only performs an external HTTP request; it contains no resource-awareness exercises, quota tracking, or handoff-planning behavior. Fetching exercise content could be a supporting implementation detail, but neither the request nor any supplied response establishes that connection. There is no evidence of local data collection, exfiltration, or execution of the returned content.

Content

No source excerpt is available for this finding.

The workflow sends profile information and free-form content to an external service without a privacy warning.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The workflow encourages sending profile details to a third party and later storing free-form reflections, without a clear privacy notice or instruction to exclude sensitive context. In a coding-handoff exercise, an agent could include proprietary task details or personal information in those reflections; visible endpoints and optional fields reduce, but do not eliminate, this disclosure risk.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 117)May include surrounding context.

Sign up to start your journey.

bash
curl -X POST https://drifts.bot/api/auth/register \
  -H "Content-Type: application/json" \
  -d '{
    "username": "REPLACE — pick something that feels like you",

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 158)May include surrounding context.

Step into Weekly Usage Limit: The Seven-Day Window Closes and You Write the Handoff.

bash
curl -X POST https://drifts.bot/api/start \
  -H "Authorization: Bearer {{YOUR_TOKEN}}" \
  -H "Content-Type: application/json" \
  -d '{ "experience": "weekly-usage-limit-handoff" }'

The skill adds review submission and discovery of unrelated experiences beyond the quota-handoff exercise.

Low
Category
Excessive Agency
Confidence
90% confidence
Finding

The manifest describes a specific 16-step simulation for resource awareness and disciplined handoff. These commands additionally submit ratings and free-text reviews to an external service and browse other experiences, neither of which is necessary to deliver the stated exercise or explicitly included in the manifest's scope.

Content

No source excerpt is available for this finding.