agent-haas

v0.3.2 safe
4.0
Medium Risk

Harness as a Service — production-grade multi-agent harness with LLM routing, paged context, hierarchical tracing, safety, and self-improvement

🤖 AI Analysis

Final verdict: SAFE

The package shows minimal risk with no clear signs of malicious activity. The high obfuscation risk is likely due to operational restrictions rather than obfuscation for malicious purposes.

  • network calls are likely legitimate
  • no shell execution detected
  • obfuscation techniques used to restrict dangerous operations
Per-check LLM notes
  • Network: The observed network calls using httpx are likely for legitimate purposes such as making HTTP requests to external services, but could be indicative of data exfiltration if the endpoints are unauthorized.
  • Shell: No shell execution patterns detected, which is normal and does not indicate any immediate risk.
  • Obfuscation: The use of regex patterns to block certain unsafe code operations like os.remove, shutil.rmtree, subprocess.call, eval, and exec suggests potential obfuscation or sandboxing but could also indicate an attempt to restrict dangerous operations.
  • Credentials: No clear patterns for harvesting credentials or secrets were detected.
  • Metadata: The maintainer has only one package, which might indicate a new or less active account, but there are no other suspicious flags.

🔬 Heuristic Checks

Outbound Network Calls score 6.0

Found 4 network call pattern(s)

  • httpx async with httpx.AsyncClient(timeout=3.0) as client: resp = await client.
  • 7474") async with httpx.AsyncClient(timeout=3.0) as client: resp = await client.
  • port httpx async with httpx.AsyncClient(timeout=5.0) as client: resp = await client.get(
  • try: async with httpx.AsyncClient(timeout=self._timeout) as client: resp = awa
Code Obfuscation score 4.0

Found 2 obfuscation pattern(s)

  • ndbox: DDL/DML not allowed in eval ({kind})", guard_source="sql_sandbox"
  • NORECASE, ) _UNSAFE_CODE = re.compile( r"\b(os\.remove|shutil\.rmtree|subprocess\.call|eval\s*\(|exec\s*\()\b", re.IGNORECASE, ) _PII = re.compile( r"(\b
Shell / Subprocess Execution

No shell execution patterns detected

Credential Harvesting

No credential harvesting patterns detected

Typosquatting

No typosquatting candidates detected

Registered Email Domain

Email domain looks legitimate: gmail.com

Suspicious Page Links

All external links appear legitimate

Git Repository History

Repository thepradip/HarnessAgent appears legitimate

Maintainer History score 2.0

1 maintainer concern(s) found

  • Author "thepradip" appears to have only 1 package on PyPI (new or inactive account)
Known CVE Vulnerabilities

No known vulnerabilities found in OSV database.

💡 AI App Starter Prompt

Use this prompt to build a project with agent-haas
Create a conversational AI assistant named 'ConversePro' that leverages the 'agent-haas' package to provide advanced conversation management capabilities. ConversePro will be designed to handle complex user queries by routing them to the most appropriate language model based on the content of the query. Additionally, it will maintain context across multiple interactions using paged context, ensure traceability through hierarchical tracing, and implement safety measures to prevent harmful outputs.

Step 1: Set up the environment by installing Python and the 'agent-haas' package.
Step 2: Design the main interface for ConversePro where users can input their queries.
Step 3: Implement a system to route user queries to different language models based on the topic and complexity of the query.
Step 4: Utilize paged context from 'agent-haas' to keep track of previous interactions and maintain continuity in conversations.
Step 5: Incorporate hierarchical tracing to log all interactions and actions taken by ConversePro for auditing purposes.
Step 6: Ensure that ConversePro includes safety mechanisms to filter out inappropriate or harmful responses.
Step 7: Allow ConversePro to improve its performance over time by implementing self-improvement features provided by 'agent-haas'.

Suggested Features:
- User authentication and personalized settings
- Integration with external APIs for enhanced functionality
- A dashboard for administrators to manage and monitor ConversePro's performance
- Support for multiple languages and dialects
- Real-time feedback mechanism for users to rate the quality of responses

The 'agent-haas' package will be utilized extensively throughout the development process to ensure that ConversePro meets the requirements for production-grade multi-agent systems, including robust LLM routing, context management, traceability, and safety.