Back to skill

Security audit

long-video-transcript

Security checks for vulnerabilities and agentic risk

Overview

The transcription service fits the skill’s purpose, but its default commands can spend more credits than the advertised discounted mode.

Before paid use, explicitly select --depth transcript and a --max-credits limit you approve. YouTube still uses full-tier billing, even when transcript mode is requested; check the [service documentation](https://linkdigest.dev/zh/docs). Submit public links only because results may be shared through caching. Prefer the bundled script or direct HTTP integration; review and pin mcp-remote if using the optional bridge.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

Tainted flow: 'req' from os.environ.get (line 102, credential/environment) → urllib.request.urlopen (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · scripts/linkdigest.py (reported line 110)May include surrounding context.

python
if data is not None:
        req.add_header("Content-Type", "application/json")
    try:
        with urllib.request.urlopen(req, timeout=timeout, context=_SSL["context"]) as resp:
            status, headers, raw = resp.status, resp.headers, resp.read()
    except urllib.error.HTTPError as err:
        status, headers, raw = err.code, err.headers, err.read()

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill promises discounted transcript mode by default, but its direct HTTP examples omit that selection, creating a credible risk of unexpected credit consumption or lower duration limits. The reported Python default cannot be independently verified without the script; optional translation and breakdown are documented and are not themselves evidence of hidden behavior, while the YouTube exception is disclosed only deeper in the document.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, network, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

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 · SKILL.md (reported line 72)May include surrounding context.

直接调 HTTP(没有 Python 时)

bash
curl -sS -X POST https://linkdigest.dev/api/v1/digest \
  -H "Authorization: Bearer $LINKDIGEST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://v.douyin.com/<短链>/", "format": "json"}'

Direct HTTP examples do not select the promised discounted long-video mode.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The direct HTTP submission specifies only url and format, despite advertising a discounted workflow that requires depth=transcript. Following this example can select the full-tier default and consume more credits than expected; this is the same underlying billing-mode issue identified in TP4, not an independent exploit.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx mcp-remote'.

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

The optional integration invokes an unversioned npm package through npx, allowing newly published package code to execute with the user's local privileges. A compromised release could access local data and the API credential supplied to the process, although no compromised package is demonstrated here.

Content

No source excerpt is available for this finding.

The default request does not select the manifest’s advertised lower-cost transcript tier.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The skill advertises the lower-cost transcript tier for long videos, but the CLI defaults to full and sends depth=transcript only when explicitly requested. An authenticated invocation using the defaults can therefore incur higher charges than advertised; the absence of a default credit ceiling makes this mismatch more consequential, although callers that explicitly select transcript avoid it.

Content

No source excerpt is available for this finding.