Back to skill

Security audit

sentisense

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious with medium confidence because scanners report an unpinned command that runs downloaded software and an initialization workflow that exceeds the declared permissions; they also flag API-key creation despite a read-only claim.

Consider the review warning before installing. The scanners recommend against installation because setup runs unpinned downloaded code and exceeds the declared permissions, and because API-key provisioning conflicts with the read-only claim. The scan was partial: although it covered the listed component, it could not resolve many local path-like references, so the evidence is incomplete.

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
  • 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
Findings (7)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Although obtaining an API key supports subsequent market-data access, the supplied command invokes a provisioning endpoint rather than a read-only data endpoint. This materially contradicts the description's explicit 'no write operations' restriction. The chunk does not show trading, purchases, or wallet access.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx sentisense'.

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The recommended npx sentisense init downloads and executes an unpinned npm package with the agent process's permissions. A compromised package release or dependency could access local files and credentials, despite the skill's otherwise narrow financial-data purpose; describing the package as official does not establish its integrity.

Content

No source excerpt is available for this finding.

The recommended CLI initialization exceeds the declared network and filesystem permissions.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The recommended CLI bootstrap uses a shell, contacts the npm distribution infrastructure, and persists an install identifier, exceeding the file's app.sentisense.ai-only network and no-files declarations. Although the exception is disclosed in prose, a host or user relying on the Permissions block could approve a much narrower capability set than initialization actually uses.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx sentisense'.

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

This is another occurrence of the same unpinned npm execution path, rather than a separate exploit mechanism. The file acknowledges that it downloads and runs code, but neither pins the executable here nor restricts the permissions that code receives.

Content

No source excerpt is available for this finding.

Temporary API-key creation contradicts the manifest's unconditional no-write guarantee.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The executable recipe at L0082 sends a POST that creates a temporary API credential, rather than merely querying market data. Although onboarding is openly documented, creating a server-side key is a write operation and conflicts with the manifest's unconditional 'Read-only' and 'no write operations' claims at L0003.

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 1742)May include surrounding context.

| limit | int | No | 100 | Rows returned, capped at 500. Sits here, not inside plan |

bash
curl -X POST "https://app.sentisense.ai/api/v1/screener/execute" \
  -H "X-SentiSense-API-Key: $SENTISENSE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{

L1759 instructs sending the user's watchlist to SentiSense without warning that this discloses potentially private financial interests to an external service.

Medium
Category
Output Handling
Confidence
87% confidence
Finding

The workflow tells the agent to include watchlist tickers in a request to SentiSense, rather than merely listing an available endpoint. Although the file identifies the API host and emphasizes read-only access, it does not warn that submitting a personal watchlist reveals that information to the service; read-only access does not prevent this privacy impact.

Content

No source excerpt is available for this finding.