MCP Registry

@observee/agents

d1a4fa52-b54d-5620-97d2-1d572a5eee26 | Store v1.0.3
100/ 100
CRITICAL risk
Risk verdict
Do not install

Score-based assessment (critical risk, 100/100). An AI review is available in the AI review tab.

Analysis record

Analysed
2 months ago 26 August 2026
Version
v1.0.3
Artifact
SHA256 5b66f227…9fd505
Detail
Full list

Evidence ledger

Ranked by severity · findings with a source location link to the code viewer

38 findings

YARA Rule Matches

8 rules
SeverityRuleHitsFilesMetadata
LOWpostinstall crypto operations 2
dist/search/bm25Filter.jsdist/search/bm25Filter.d.ts
-
LOWpostinstall file manipulation 2
dist/search/bm25Filter.jsdist/providers/anthropic.js
-
LOWpostinstall registry modification 3
dist/providers/index.jsdist/agents/agent.jsdist/providers/index.d.ts
-
LOWpostinstall obfuscation 2
dist/adapters/langchain-adapter.jsREADME.md
-
LOWpostinstall network communication 4
dist/agents/agent.jsREADME.mddist/utils/mcp-client.js +1 more
-
LOWpostinstall system command 9
dist/index.d.tsdist/utils/mcp-client.jsdist/types.d.ts +6 more
-
LOWpostinstall file download 1
dist/utils/mcp-client.js
-
LOWcredential env files 5
dist/agents/agent.jsdist/index.jsdist/providers/openai.js +2 more
-

Finding Categories

10
Secrets
15
IoC Indicators

YARA Rules Matched

8 rules(28 hits)
postinstall crypto operations postinstall file manipulation postinstall registry modification postinstall obfuscation postinstall network communication postinstall system command postinstall file download credential env files

MCP Server Analysis

MCP servers expose tools and resources to AI assistants. Unlike browser extensions, they run as standalone processes with direct system access. Tool definitions are analyzed for prompt injection, data exfiltration, and tool poisoning patterns.

AI Security Report

AI Security Review

Evidence context: threat category unknown malware; evidence quality weak.

This package presents a critical risk score driven by a high volume of IoC and malware-signature findings, yet the provided evidence bundle lacks the specific file paths, code snippets, or domain details required to validate these alerts. In the MCP ecosystem, high finding counts are frequently false positives caused by bundled dependencies in dist/ files; a single bundled library can trigger hundreds of benign IoCs (e.g., CDN domains) and generic YARA matches. While the scanner's confidence (0.9) and CRITICAL categorization suggest a severe anomaly, the absence of specific tool-poisoning evidence (the defining MCP threat) or clear exfiltration patterns (credential access paired with unknown network destinations) makes it impossible to confirm malicious intent from this summary alone. This appears to be a 'noisy' package that requires a detailed file-by-file review to distinguish between a supply-chain threat and a false positive from an aggressive scanner.

Key Reasons

  • CRITICAL risk score (100) flagged with zero user base adoption
  • High volume of IoC (73) and Malware-signature (28) findings detected
  • No tool-poisoning findings present despite critical severity
  • Missing specific finding details (domains, file paths, code snippets)

False Positive Considerations

  • Bundled dependencies in dist/ files inflating IoC counts
  • Generic YARA malware-signature rules matching common code patterns

Reviewed 2026-04-13; recommended action: reanalyze; model confidence 50%.

Source Code Viewer

Source Code Not Available

Source code is not available for this version of the extension.

About This Extension

Observee SDK - A TypeScript SDK for MCP tool integration with LLM providers

Frequently Asked Questions