Back to skill

Security audit

searchtok

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent SearchTok social-video research helper, but it repeatedly tells agents to run mutable unpinned npx commands, including auth and scheduled workflows, which should be reviewed before installation.

Install only if you trust SearchTok and are comfortable giving its MCP server or CLI access to your SearchTok account and research context. Prefer the MCP connector or a pinned, reviewed CLI installation over running bare `npx searchtok`, and review any scheduled monitoring task before enabling it.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (67)

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · references/brand-analysis.md (reported line 55)May include surrounding context.

md
## Don'ts

- Don't score a post without its numbers, and use public numbers only (views, likes, comments, shares, saves).
- Don't judge on conversions: searchtok can't see them.
- Don't rank a brand's posts by raw views: judge them against the brand's usual views.
- Never show low-view videos, and never say why a video didn't work.
- Leave ads out entirely: never show one or use it as evidence.

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · references/build-profile.md (reported line 56)May include surrounding context.

md
## Don'ts

- Don't score a post without its numbers, and use public numbers only (views, likes, comments, shares, saves).
- Don't judge on conversions: searchtok can't see them.
- Don't rank a brand's posts by raw views: judge them against the brand's usual views.
- Never show low-view videos, and never say why a video didn't work.
- Leave ads out entirely: never show one or use it as evidence.

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · references/content-ideas.md (reported line 60)May include surrounding context.

md
## Don'ts

- Don't score a post without its numbers, and use public numbers only (views, likes, comments, shares, saves).
- Don't judge on conversions: searchtok can't see them.
- Don't rank a brand's posts by raw views: judge them against the brand's usual views.
- Never show low-view videos, and never say why a video didn't work.
- Leave ads out entirely: never show one or use it as evidence.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The invocation guidance says to use the skill when someone asks "what's working, trending or blowing up" or wants "content ideas or a script structure," which are broad everyday phrases not clearly limited to TikTok/Instagram research. Because the trigger list is expansive and lacks negative examples or tighter activation boundaries, it could cause unintended invocation for generic marketing or writing requests.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The skill repeatedly instructs use of npx searchtok without pinning an exact package version. npx resolves and executes the latest package version at runtime, so a compromised publisher account, dependency-chain attack, or malicious new release could cause arbitrary code execution in the agent environment.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding

The same line also instructs users to run npx searchtok auth:login, again invoking whatever package version npm currently resolves. That expands the attack surface to browser-based auth flows and local credential handling under potentially untrusted code.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The free usage-check instruction references npx searchtok usage without version pinning. Even seemingly low-risk status commands still execute fetched code, so an attacker could exploit the trust implied by a harmless-looking command to run arbitrary logic.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The search_content CLI example uses unpinned npx searchtok search, allowing arbitrary code execution if the upstream package or dependencies are compromised. Because this command is central to normal operation, users are likely to execute it frequently, increasing exposure.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The find_accounts CLI example is unpinned and therefore inherits the same npx supply-chain execution risk. Frequent use in agent workflows makes this more dangerous because the skill normalizes executing remote package code as part of routine research.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The get_account example executes npx searchtok account without constraining the version. A malicious or typosquatted package update could harvest environment data, modify outputs, or run arbitrary commands under the agent's identity.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The analyze_brand command is unpinned and may process user-provided URLs, handles, or names using code fetched at runtime. That combination of dynamic execution plus external input increases risk if the package becomes malicious or compromised.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The analyze_post example again relies on npx without a fixed version, creating a repeatable supply-chain path to execute arbitrary code. Since this skill encourages passing URLs and IDs directly into commands, compromised tooling could also exfiltrate those inputs or related auth material.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The find_similar CLI example is another unpinned runtime package execution point. Repetition throughout the skill makes the issue systemic rather than isolated, raising the likelihood that an operator will follow unsafe examples verbatim.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The build_profile command uses unpinned npx, which can execute whatever version npm serves at invocation time. Because this command may ingest business descriptions and links, a compromised package could collect sensitive context or manipulate profile-building outputs.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The get_profile CLI example is unpinned and could expose user profile information to a compromised package. Even read-only looking commands are dangerous because the core risk is code execution, not the nominal business function.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The get_usage CLI example continues the pattern of unpinned npx execution. Attackers often rely on low-suspicion helper commands to gain execution because users assume they are safe administrative checks.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The research command invokes a high-value workflow via unpinned npx, increasing both execution frequency and the amount of context potentially exposed to compromised code. Because it may run broad research tasks, it could handle substantial user data and network activity.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The get_research example is unpinned and therefore unsafe for the same supply-chain reasons as the other commands. The command may retrieve prior research artifacts, so compromised code could tamper with or exfiltrate stored analysis results.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The content_ideas command is another unpinned executable fetch. Since it may operate on saved profiles and generate user-targeted strategy outputs, compromise could leak profile data or poison business recommendations at scale.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The same line also recommends npx searchtok similar without version pinning, creating a second unsafe execution path. Multiple unpinned commands on one line reinforce insecure defaults and make accidental exposure more likely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The instruction "Default to the US and Europe" imposes a locale choice automatically rather than asking the user or offering alternatives. This is a natural-language locale policy issue because it forces a regional default unless the user explicitly overrides it.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The recipe instructs users to run npx searchtok ... without pinning a specific package version, so execution depends on whatever version is currently published or resolved at runtime. If the package is updated maliciously or a compromised dependency is served, scheduled or repeated executions could run unreviewed code on the user's machine or agent environment.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This second npx searchtok post <url> invocation has the same supply-chain risk: it fetches or resolves a potentially changing package at runtime instead of a reviewed fixed version. In this skill, the risk is amplified because the command is embedded in an automation recipe, increasing the chance of unattended recurring execution of compromised code.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The recipe instructs users to run npx searchtok without pinning an exact package version, so execution depends on whatever version is currently published or resolved at runtime. If the package, a dependency, or the name resolution path is compromised, scheduled or repeated runs could fetch and execute attacker-controlled code on the user's machine or automation environment.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This command again uses npx searchtok without a pinned version, but here it is applied to per-post analysis and may be run repeatedly as part of the workflow. In the skill context, the recipe encourages recurring use and automation, which increases exposure to supply-chain compromise because each execution can resolve and run unpinned remote package code.

Content

No source excerpt is available for this finding.