Back to skill

Security audit

Tick-Coord: Multi-Agent Task Coordination

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches its task-coordination purpose, but its dashboard instructions can expose project task data over the network and sync it to a remote service, so it needs review before installation.

Install only if you are comfortable with agents maintaining TICK.md task state and, if using the dashboard, exposing that state through a temporary web server. Prefer local-only operation, bind the dashboard to 127.0.0.1, stop it after use, avoid syncing sensitive task data to Convex unless you control the deployment and key, and review the referenced scripts before running them.

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
SKILL.md:498
Finding

Unauthenticated Development Dashboard Exposed on All Network Interfaces

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 498-532
Vulnerability Type: Unauthenticated network exposure of sensitive project data
Risk Level: High

Vulnerable Code

bash
# CORRECT — detached, survives exec timeout
cd "$SKILL_DIR/web"
setsid pnpm dev --port 3000 --host 0.0.0.0 </dev/null > /tmp/tick-board.log 2>&1 &
echo $! > /tmp/tick-board.pid

The accompanying instructions explicitly state:

text
The dev server binds to `0.0.0.0`, making it available on all network interfaces.

They also acknowledge that the development server has no built-in authentication or encryption and that task data may contain sensitive project information.

Technical Analysis

Binding the dashboard to 0.0.0.0 exposes it through every available network interface rather than limiting access to the local machine. The server is also launched as a detached background process, increasing the likelihood that it will continue running after the user finishes inspecting the board.

The dashboard is described as read-only, but it displays project lists, task details, task history, priorities, tags, and coordination data. Such records can contain confidential project plans, internal endpoints, implementation decisions, employee or agent identities, and security-related notes.

The documentation includes warnings against public exposure, but these warnings do not correct the insecure default command. Local dashboard inspection only requires binding to 127.0.0.1; therefore, exposing all interfaces exceeds the minimum network privileges necessary for the declared functionality.

The referenced scripts/tick-board.sh implementation is not included in the audited package, so it was not possible to verify whether that script adds authentication, restricts network binding, validates ports, or reliably terminates stale processes.

Attack Path

  1. A user follows the Skill instructions ...[truncated 1391 chars]
Remediation
View remediation

Remediation Suggestions

  1. Bind to the loopback interface by default:

    bash
    setsid pnpm dev --port 3000 --host 127.0.0.1 </dev/null > /tmp/tick-board.log 2>&1 &
    
  2. Require explicit user confirmation and a separate opt-in flag before binding to a LAN, VPN, or public-facing interface.

  3. For remote access, place the dashboard behind an authenticated reverse proxy with TLS rather than exposing the development server directly.

  4. Add startup checks that detect public-facing interfaces, permissive firewall rules, occupied ports, and stale PID files.

  5. Implement automatic shutdown after an inactivity timeout or at the end of the inspection session.

  6. Ensure the stop operation validates that the PID belongs to the expected dashboard process before terminating it.

  7. Avoid placing credentials or sensitive data in process arguments, URLs, or logs.

  8. Include and audit scripts/tick-board.sh in the distributed package so its binding, authentication, process-management, and cleanup behavior can be verified.

  9. Clearly separate safe local operation from remote deployment guidance; remote deployment should require authentication, encrypted transport, and restrictive network access controls.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (6)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill’s declared purpose is local Git-backed Markdown coordination, but this section expands behavior into launching and using a networked dashboard plus syncing data to a remote Convex service. That broadens the trust boundary from local files to network services and creates opportunities for unintended data exposure or unauthorized remote access that are not inherent to the core skill.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

Broad trigger language such as 'show me the board' or 'or similar' can cause the agent to launch the dashboard in response to ambiguous requests that might only mean 'summarize project status.' In this skill, that ambiguity is more dangerous because the action can start a network service rather than merely display information.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The instructions explicitly start a detached dev server bound to 0.0.0.0, making it reachable from other hosts while also causing it to persist beyond the initiating command session. A persistent, network-accessible service with no built-in authentication materially increases exposure, especially in agent contexts where users may not realize a long-lived server has been started.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
96% confidence
Finding

Using setsid to launch the dev server as a detached background process creates session persistence outside the normal execution lifecycle. In combination with 0.0.0.0 binding and lack of authentication, this can leave an unnoticed service running indefinitely, increasing the chance of accidental exposure and making cleanup harder.

Content

Scanner excerpt · SKILL.md (reported line 503)May include surrounding context.

bash
# CORRECT — detached, survives exec timeout
cd "$SKILL_DIR/web"
setsid pnpm dev --port 3000 --host 0.0.0.0 </dev/null > /tmp/tick-board.log 2>&1 &
echo $! > /tmp/tick-board.pid

# WRONG — will be killed when exec times out

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The documentation says the dashboard is read-only, but the adjacent workflow instructs syncing task data to a remote Convex deployment using credentials. Even if the UI itself is read-only, the skill still enables outbound state publication, which can mislead users about confidentiality and system effects.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The sync workflow sends task data to a remote Convex deployment, but the workflow step itself does not present a direct user-facing warning or consent check at the moment of action. That creates a risk of unintentional exfiltration of potentially sensitive task content, metadata, comments, or project structure.

Content

No source excerpt is available for this finding.