Back to skill

Security audit

Speech to Text

Security checks for vulnerabilities and agentic risk

Overview

This transcription skill is mostly transparent about uploading audio to a public Whisper service, but its helper script can upload any readable local file to any supplied endpoint if misused.

Install only if you are comfortable with selected audio being uploaded to a public third-party Whisper/Gradio service. Before running it, verify the exact local file path and remote endpoint, avoid sensitive recordings, and do not let untrusted text choose the input path or --space value.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/transcribe.py:23
Finding

Arbitrary Local File Upload to an Unrestricted Remote Endpoint

Content
View full analysis
str: boundary = "----OpenClawBoundary" + uuid.uuid4().hex mime = mimetypes.guess_type(str(path))[0] or "application/octet-stream" body = [] body.append(f"--{boundary}\r\n".encode()) body.append( f'Content-Disposition: form-data; name="files"; filename="{path.name}"\r\n'.encode() ) body.append(f"Content-Type: {mime}\r\n\r\n".encode()) body.append(path.read_bytes()) body.append(f"\r\n--{boundary}--\r\n".encode()) data = b"".join(body) req = request.Request( f"{space}/gradio_api/upload", data=data, method="POST", headers={ "Content-Type": f"multipart/form-data; boundary={boundary}", "User-Agent": USER_AGENT, }, ) with request.urlopen(req, timeout=300) as resp: uploaded = json.loads(resp.read().decode("utf-8")) ``` ```python def transcribe_file(path: Path, space: str, task: str) -> str: if not path.exists(): raise FileNotFoundError(path) mime = mimetypes.guess_type(str(path))[0] or "application/octet-stream" server_path = upload_file(space, path) event_id = submit_job(space, server_path, path.name, mime, task) return wait_for_result(space, event_id) ``` ```python parser.add_argument("input", help="Path to local audio file") parser.add_argument("--task", choices=["transcribe", "translate"], default="transcribe") parser.add_argument("--space", default=DEFAULT_SPACE, help="Base URL of the Gradio space") parser.add_argument("--format", choices=["auto", "raw", "zh"], default="auto", help="Post-process transcript punctuation") parser.add_argument("--json", action="store_tru ...[truncated 2846 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill instructs the agent to read local audio files, access environment configuration, and send file contents to a public Hugging Face/Gradio endpoint, but it declares no explicit tool scope or permissions. This creates a real security gap because an agent or platform may allow broader-than-intended file, network, or environment access without clear consent boundaries, especially given the documented override of the remote endpoint via CLI flag or environment variable.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The script automatically applies Chinese-specific punctuation rules whenever looks_like_chinese returns true, and it also offers a zh mode that enforces this locale-specific formatting. This is a language/locale behavior choice embedded in the skill without an explicit opt-in prompt or documented justification in the file.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.