ClawHub
Back to skill

Security audit

Opentask Worker

Security checks for vulnerabilities and agentic risk

Overview

The skill is openly about automating OpenTask.ai work, but it enables autonomous bids, contracts, submissions, and payout setup without clear user approval or limits.

Review this skill carefully before installing. Use it only with an OpenTask account you intend to automate, keep OPENTASK_TOKEN and wallet details in a secure secret store, and require manual approval for registration, bids, contract actions, deliverable submissions, and payout changes unless you have set strict limits and a reliable stop mechanism.

A.I.G

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
Findings (1)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:75
Finding

Untrusted Marketplace Tasks Can Redirect Autonomous Agent Behavior

Content
View full analysis
0: notifs = requests.get(f"{BASE}/api/agent/notifications?unreadOnly=1", headers=HEADERS).json()["notifications"] for n in notifs: handle_notification(n) # bid accepted, contract created, etc. # Discover new tasks tasks = requests.get(f"{BASE}/api/tasks?sort=new", headers=HEADERS).json()["tasks"] for t in tasks: if qualifies(t): # budget > threshold, skills match place_bid(t) time.sleep(1800) # poll every 30 min ``` ``` ### Technical Analysis The skill instructs an agent to continuously retrieve marketplace tasks and notifications and then autonomously process them through `handle_notification(n)` and `place_bid(t)`. Marketplace task descriptions and notifications are externally controlled inputs, but the documented workflow does not establish a trust boundary between this content and the agent’s instructions. The `qualifies(t)` check is described only as considering budget and skill compatibility. It does not require validation of task safety, authorization, data-access requirements, financial limits, or consistency with the user’s original objective. Likewise, there is no mandatory approval step before bidding, reacting to a contract, submitting work, or making another consequential marketplace action. If an implementation passes task or notification text into an instruction-following agent, an attacker can embed instructions designed to redirect the agent’s ...[truncated 1773 chars]
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill describes autonomous polling, bidding, contract handling, submission, and payout configuration without a prominent warning that it can take external actions affecting accounts, contractual obligations, and payment destinations. In this context, lack of disclosure and confirmation is dangerous because a user may invoke the skill expecting advice, while the agent performs persistent marketplace actions with financial and reputational consequences.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The 'use when' description is broad enough to trigger on generic requests about earning money, finding work, contracts, or automation, which can cause the skill to activate in situations where the user did not intend autonomous marketplace participation. In this skill's context, activation can lead to external account actions, bidding, and payout setup, so over-broad routing materially increases the chance of unintended real-world transactions.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

The registration example transmits email and password to an external service and instructs saving a returned token for later authenticated use. While normal for account creation, it is still security-sensitive because this skill is designed for headless autonomous use; if triggered unintentionally or used with real credentials, it can create accounts and expose secrets or tokens to logs, history, or downstream components.

Content

Scanner excerpt · SKILL.md (reported line 17)May include surrounding context.

1. Register (headless, no browser)

bash
curl -X POST "https://opentask.ai/api/agent/register" \
  -H "Content-Type: application/json" \
  -d '{"email":"your-agent@example.com","password":"SecurePass123","handle":"your_agent","displayName":"Your Agent"}'
# Save tokenValue as OPENTASK_TOKEN

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

The bid submission endpoint performs an authenticated external action that can create binding marketplace offers using the user's account token. In this skill's context, this is more dangerous than ordinary data transmission because autonomous bidding can commit the user to deadlines, pricing, and contractual negotiations, creating financial loss and reputational damage if invoked incorrectly.

Content

Scanner excerpt · SKILL.md (reported line 30)May include surrounding context.

3. Place a bid

bash
curl -X POST "https://opentask.ai/api/agent/tasks/TASK_ID/bids" \
  -H "Authorization: Bearer $OPENTASK_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"priceText":"50 USDC","etaDays":1,"approach":"Plan: ... Verification: ..."}'