ClawHub
Back to skill

Security audit

tavily-search-pro-native-node

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed Tavily web research helper whose network use, API key use, local cache, logs, and cache deletion are coherent with its purpose.

Install only if you are comfortable sending approved queries and URLs to Tavily with your TAVILY_API_KEY. Use --no-log --no-cache for sensitive or client/private research, use separate OS profiles when Tavily account isolation matters, and only run cache clear when you intentionally want to delete this skill's cached response files.

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 (2)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/tavily-pro.mjs:320
Finding

Plaintext Search Queries and Extraction URLs Are Logged by Default

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
scripts/tavily-pro.mjs:773
Finding

Documented Approval Gate for Cache Deletion Is Not Enforced by the CLI

Content
View full analysis
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 104)May include surrounding context.

md
Script: `scripts/tavily-pro.mjs`

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/self-test.mjs (reported line 42)May include surrounding context.

js
function run(name, args, { env = {}, expectCode = 0, stderrIncludes, stdoutIncludes, stderrExcludes, stdoutExcludes } = {}) {
  const result = spawnSync(process.execPath, [cli, ...args], {
    env: { ...baseEnv, ...env },
    encoding: "utf8",
  });
  const stdout = result.stdout || "";

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
84% confidence
Finding

The skill declares access to environment variables and performs outbound network calls, but it does not define an explicit tool scope such as permissions or allowed-tools. In an agent ecosystem, missing capability declarations weaken policy enforcement and user visibility, increasing the chance that a skill with external transmission behavior is invoked without appropriate guardrails.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
60% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · SKILL.md (reported line 74)May include surrounding context.

md
- Usage logs may contain plaintext search queries/URLs; use `--no-log` for approved sensitive external calls.
- Local/private URL refusal is a guardrail for obvious mistaken extract targets, not a complete SSRF boundary; Tavily performs extraction from Tavily infrastructure, not this machine.
- Does not read local files outside its documented cache/log state, and does not read non-Tavily secrets. User-supplied search queries and accepted extract URLs are sent to Tavily; `TAVILY_API_KEY` is sent only to Tavily for authentication. Credential-bearing and local/private extract URLs are rejected before API key load/Tavily transmission, and those error paths do not echo the supplied URL.
- Does not modify system configuration or auto-update. The only destructive local action is the documented, approval-gated `cache clear` deletion of expected regular cached response JSON files after cache-path containment checks.
- Public-registry static-analysis `potential_exfiltration` warnings are expected because this tool combines env credentials, local cache/log file access, and Tavily network calls.

## When to use

Session Persistence

Medium
Category
Rogue Agent
Confidence
94% confidence
Finding

The contract explicitly documents that cached Tavily responses are not scoped to the API account and may be shared when multiple Tavily accounts use the same OS user/home directory. This can expose prior search/extract results across account boundaries, which is a confidentiality and session-persistence issue, especially because the skill handles research queries and extracted web content that may be sensitive.

Content

Scanner excerpt · references/tavily-pro-contract.md (reported line 80)May include surrounding context.

md
`cache` with no subcommand defaults to `cache info`. Routine cache reads/writes and usage-log reads/writes perform persistent-state containment checks and refuse symlink-indirected directories/files instead of following them outside `~/.openclaw/cache/tavily-search-pro-native-node/`. `cache clear` accepts no extra arguments, requires explicit approval, rejects symlink-indirected cache paths, deletes only expected hash-named regular cached response JSON files under the skill-specific cache directory, skips non-cache entries, and does not delete `usage.log`. It reports partial deletion failures with the number already deleted instead of returning a false zero.

Responses are cached under `~/.openclaw/cache/tavily-search-pro-native-node/cache/` keyed by a SHA-256-derived hash of request body + kind. Request-body-affecting options such as topic, depth, max results, include/exclude domains, raw-content, URL list, and extract depth create different cache entries; output/control flags such as `--json`, `--ttl`, `--no-log`, and `--no-retry` do not. Cache entries are not API-account-scoped; if multiple Tavily accounts share the same OS user/home directory, they may share cached results for identical requests. Use separate profiles or `--no-cache` when account isolation matters.

Default TTLs:

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 66)May include surrounding context.

md
import { homedir } from "node:os";
import { join, relative, resolve } from "node:path";

const SEARCH_ENDPOINT = "https://api.tavily.com/search";
const EXTRACT_ENDPOINT = "https://api.tavily.com/extract";
const SKILL_DIR_NAME = "tavily-search-pro-native-node";
const BASE_DIR = join(homedir(), ".openclaw", "cache", SKILL_DIR_NAME);

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 67)May include surrounding context.

md
import { homedir } from "node:os";
import { join, relative, resolve } from "node:path";

const SEARCH_ENDPOINT = "https://api.tavily.com/search";
const EXTRACT_ENDPOINT = "https://api.tavily.com/extract";
const SKILL_DIR_NAME = "tavily-search-pro-native-node";
const BASE_DIR = join(homedir(), ".openclaw", "cache", SKILL_DIR_NAME);

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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/tavily-pro.mjs (reported line 12)May include surrounding context.

js
import { homedir } from "node:os";
import { join, relative, resolve } from "node:path";

const SEARCH_ENDPOINT = "https://api.tavily.com/search";
const EXTRACT_ENDPOINT = "https://api.tavily.com/extract";
const SKILL_DIR_NAME = "tavily-search-pro-native-node";
const BASE_DIR = join(homedir(), ".openclaw", "cache", SKILL_DIR_NAME);

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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/tavily-pro.mjs (reported line 13)May include surrounding context.

js
import { homedir } from "node:os";
import { join, relative, resolve } from "node:path";

const SEARCH_ENDPOINT = "https://api.tavily.com/search";
const EXTRACT_ENDPOINT = "https://api.tavily.com/extract";
const SKILL_DIR_NAME = "tavily-search-pro-native-node";
const BASE_DIR = join(homedir(), ".openclaw", "cache", SKILL_DIR_NAME);

Static analysis

Detected: suspicious.dangerous_exec, suspicious.env_credential_access

Shell command execution detected (child_process).

Critical
Code
suspicious.dangerous_exec
Location
scripts/self-test.mjs:41

Environment variable access combined with network send.

Critical
Code
suspicious.env_credential_access
Location
scripts/tavily-pro.mjs:43