## PraisonAI Codebase Exposes 29 Shell Injection Vulnerabilities (CWE-78) Post-Audit
A critical security audit of the PraisonAI codebase has revealed 29 unaddressed shell injection vulnerabilities, classified as CWE-78, posing a direct risk of arbitrary command execution. These high-risk flaws persist despite a recent security push that successfully patched other issues, indicating a deliberate deferral that leaves the system's security perimeter dangerously porous.

The vulnerabilities are concentrated in the use of Python's `subprocess` module. Fourteen high-risk instances involve passing unsanitized user input directly to `subprocess.run` calls. Another ten high-risk cases stem from the use of `shell=True` without proper safeguards, while five medium-risk instances involve tainted environment arguments. Collectively, these flaws create multiple vectors where an attacker could inject and execute malicious shell commands, potentially leading to full system compromise.

The discovery follows a prior audit (PR #1319) that fixed 29 MD5 and debug-related vulnerabilities but explicitly excluded these shell injection issues, citing a need for architectural decisions from project maintainers. This bifurcation suggests internal conflict or uncertainty over remediation strategy. The unresolved status of these high-severity findings, confirmed by semgrep scans using OWASP Top 10 rulesets, places ongoing pressure on the PraisonAI team to prioritize a fundamental security overhaul over incremental fixes.
---
- **Source**: GitHub Issues
- **Sector**: The Lab
- **Tags**: Security Vulnerability, Code Audit, Shell Injection, CWE-78, Python
- **Credibility**: unverified
- **Published**: 2026-04-09 06:27:11
- **ID**: 56319
- **URL**: https://whisperx.ai/en/intel/56319