ClawHub
Back to skill

Security audit

openai-whisper-api

Security checks for vulnerabilities and agentic risk

Overview

This skill does what it claims, but its custom API endpoint setting can send both the user's audio and OpenAI API key to an arbitrary configured URL without validation.

Install only if you understand that selected recordings and prompt text will be uploaded to the configured transcription endpoint. Avoid using OPENAI_BASE_URL unless you fully trust that gateway, and do not point it at HTTP or third-party endpoints with your primary OpenAI API key.

A.I.G

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

Error
Location
scripts/transcribe.sh:69
Finding

Unrestricted API Base URL Can Expose the API Key and Audio Data

Content
View full analysis
&2 exit 2 fi target="$out" tmp="" if [[ "$json_output" == "0" && ( "$request_format" == "json" || "$request_format" == "diarized_json" ) ]]; then tmp="$(mktemp)" trap '[[ "$tmp" == "" ]] || rm -f "$tmp"' EXIT target="$tmp" fi curl_args=( -sS "${api_base}/audio/transcriptions" -H "Authorization: Bearer $OPENAI_API_KEY" -H "Accept: application/json" -F "file=@${in}" -F "model=${model}" -F "response_format=${request_format}" ) ``` The corresponding configurable-proxy behavior is documented in `SKILL.md` at lines 29 and 58. ### Technical Analysis The script obtains the transcription endpoint directly from the environment variable `OPENAI_BASE_URL`. It removes a trailing slash but performs no validation of the URL scheme, destination host, port, or embedded credentials. The resulting endpoint is passed to `curl` together with two sensitive assets: 1. `OPENAI_API_KEY`, transmitted in the `Authorization` header. 2. The complete user-selected audio file, transmitted as multipart form data. Consequently, any party capable of controlling the process environment or active OpenClaw configuration can redirect the request to an arbitrary server. The implementation also permits plaintext `http://` endpoints, ...[truncated 1897 chars]
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
90% confidence
Finding

Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Content

Scanner excerpt · SKILL.md (reported line 44)May include surrounding context.

Useful flags

bash
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-transcribe --out /tmp/transcript.txt
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-mini-transcribe
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-transcribe-diarize --json
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model whisper-1

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
90% confidence
Finding

Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Content

Scanner excerpt · SKILL.md (reported line 45)May include surrounding context.

bash
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-transcribe --out /tmp/transcript.txt
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-mini-transcribe
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-transcribe-diarize --json
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model whisper-1
{baseDir}/scripts/transcribe.sh /path/to/audio.m4a --language en

External Model or Provider Selection

High
Category
Excessive Agency
Confidence
90% confidence
Finding

Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Content

Scanner excerpt · SKILL.md (reported line 46)May include surrounding context.

bash
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-transcribe --out /tmp/transcript.txt
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-mini-transcribe
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model gpt-4o-transcribe-diarize --json
{baseDir}/scripts/transcribe.sh /path/to/audio.ogg --model whisper-1
{baseDir}/scripts/transcribe.sh /path/to/audio.m4a --language en
{baseDir}/scripts/transcribe.sh /path/to/audio.m4a --prompt "Speaker names: Peter, Daniel"

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding

The skill documents shell-based execution (scripts/transcribe.sh, curl, node) but does not declare an explicit tool scope such as permissions or allowed-tools. That omission can cause users or orchestrators to underestimate the skill's ability to execute commands and transmit local files, increasing the risk of unintended shell access and data handling without clear consent boundaries.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill interfaces with a remote transcription endpoint and accepts both local audio and optional prompt text, but the description does not clearly warn that these inputs are uploaded off-host. Users may provide sensitive recordings, names, or contextual prompts under the false assumption that processing is local, leading to unintended disclosure of private or regulated data.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · scripts/transcribe.sh (reported line 75)May include surrounding context.

sh
mkdir -p "$(dirname "$out")"

api_base="${OPENAI_BASE_URL:-https://api.openai.com/v1}"
api_base="${api_base%/}"

request_format="text"

Static analysis

No suspicious patterns detected.