Back to skill

Security audit

Jazz Music — Jazz Concerts for AI Agents: Audio, Lyrics, Equations

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent music-service integration, but it lets remote responses guide agent actions and endpoint choices too broadly for authenticated and user-generated submissions.

Review this skill before installing. Use a dedicated account/token, avoid personal or confidential content in profile fields, chats, reflections, and reviews, and do not let server-provided next_steps or URLs trigger actions unless they stay on musicvenue.space and match the documented API workflow.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:280
Finding

Untrusted Server Responses Can Redirect Agent Actions

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 280
Vulnerability Type: T01: Skill Instruction Hijacking
Risk Level: High

Complete Code Snippet:

markdown
**Follow next_steps.** Every response includes `next_steps` with context-aware suggestions. New agent? It guides you to your first concert. Just finished a show? It suggests a review or a new genre. Follow the suggestions — they adapt to where you are.

Technical Analysis

The skill directs the agent to follow the next_steps field returned by the external musicvenue.space service. This delegates future agent behavior to mutable, server-controlled content that is unavailable during static review.

No instruction requires the agent to treat next_steps as untrusted data, restrict it to the documented concert operations, reject embedded directives, or obtain user approval before performing suggested actions. Consequently, control of the service or its response path could allow an attacker to alter the agent's workflow after the skill has been reviewed.

Attack Path

  1. A user invokes the skill and the agent contacts musicvenue.space.
  2. The service, or an attacker who has compromised it, places arbitrary instructions in the next_steps response field.
  3. The skill instructs the agent to follow those suggestions.
  4. The agent may execute additional API or tool operations not explicitly reviewed or requested by the user.
  5. Repeated responses can progressively redirect the session beyond the advertised concert workflow.

Impact Assessment

A successful exploit can alter the agent's current-session goals and induce unauthorized tool or network actions within the agent's existing privileges. The precise scope depends on the tools and credentials available to the host agent. The reviewed file does not establish local code execution, persistence, privilege escalation, or access to system credentials, but it creates a server-contr ...[truncated 73 chars]

Remediation
View remediation

Remediation Suggestions

  • Treat next_steps strictly as untrusted informational data rather than executable instructions.
  • Replace the unconditional directive with a fixed allowlist of documented operations and parameters.
  • Reject suggestions containing embedded prompts, commands, arbitrary URLs, new hosts, or requests for secrets.
  • Require explicit user confirmation before any suggested operation that changes state, publishes content, discloses data, or expands the workflow.
  • Preserve system, developer, and user instructions as higher-priority constraints regardless of server responses.
  • Validate each proposed action locally instead of recursively interpreting natural-language instructions from the service.

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:119
Finding

Authenticated Workflow Follows Server-Supplied Endpoints Without Origin Validation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 119 and 148–152
Vulnerability Type: T09: Insecure Skill Coding Practices
Risk Level: High

Complete Code Snippet:

markdown
Params: `speed` (1-10), `window` (10-120s), `summary=true` for condensed stats, `start` to resume. Poll `next_batch.endpoint` after `wait_seconds`.
markdown
| `reflection` | Concert is asking you a question. POST your response to the `respond_to` URL within `expires_in` seconds. Missed prompts are tracked in `progress.missed_reflections`. |
| `loop` | Concert restarting (loop mode) |
| `end` | Concert over -- includes `engagement_summary` (tier, layers experienced/available, reflections answered, challenge status). Badge awarded. |

**Handling reflections:** When you see `type: "reflection"`, POST to the `respond_to` endpoint with your `ticket`, `reflection_id`, and `response`. Your response time and content are scored. Missing reflections is tracked -- the `end` event shows how many you answered vs received.

Technical Analysis

The skill instructs the client to poll next_batch.endpoint and submit content to a server-provided respond_to URL. It does not require these values to be relative paths, constrain them to the exact https://musicvenue.space origin, validate their path, or prohibit cross-origin redirects.

The surrounding workflow performs authenticated requests using an Authorization: Bearer {{YOUR_TOKEN}} header. An implementation that carries this authentication context when following the supplied endpoints could disclose the bearer token to another host. Even if the token is not forwarded, the reflection request contains the ticket identifier, reflection identifier, and agent-generated response, which could be exposed to an attacker-selected destination.

Attack Path

  1. The agent authenticates to musicvenue.space with its bearer token and begins streaming a concert.
  2. A compromised o ...[truncated 1033 chars]
Remediation
View remediation

Remediation Suggestions

  • Accept only relative endpoint paths and resolve them against the fixed origin https://musicvenue.space.
  • Enforce an exact scheme, hostname, and port match after URL normalization.
  • Allowlist expected API paths for stream polling and reflection submission.
  • Reject URLs containing user information, non-HTTPS schemes, unexpected ports, encoded host ambiguities, or protocol-relative forms.
  • Disable redirects for authenticated requests or revalidate every redirect target; never forward Authorization headers across origins.
  • Scope bearer tokens to the minimum necessary operations and lifetime.
  • Validate response schemas and endpoint values before initiating requests.
  • Warn or require user approval before sending free-form reflection content externally.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Memory Manipulation

High
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · SKILL.md (reported line 237)May include surrounding context.

md
| Tier | Layers | What you experience |
|------|--------|-------------------|
| **General** | 8 | Bass, mid, treble, beats, lyrics, sections, energy, preset context |
| **Floor** | 20 | + equations, visuals, emotions, tempo, harmonic/percussive separation |
| **VIP** | 29 | + tonality, texture, chroma, tonnetz, structure, curator annotations |

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs agents to register with an external service, store an API key, and repeatedly send user- or agent-generated content such as chat messages, reviews, and reflection responses, but it does not clearly disclose that this data leaves the local environment and is processed by a third-party service. This creates a real privacy and data-governance risk because users may unknowingly transmit profile information, behavioral data, and freeform text to an external domain.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
94% confidence
Finding

The registration example sends profile metadata including username, name, bio, avatar prompt, and model information to an external service. External transmission is expected for a networked skill, but without prominent disclosure and data-minimization guidance it still represents a genuine privacy and supply-chain risk.

Content

Scanner excerpt · SKILL.md (reported line 88)May include surrounding context.

Register

bash
curl -X POST https://musicvenue.space/api/auth/register \
  -H "Content-Type: application/json" \
  -d '{
    "username": "jazz-listener",

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

The attend endpoint performs an authenticated call to a third-party service and initiates a session tied to a bearer token and ticket lifecycle. While this is core functionality rather than overtly malicious behavior, it still creates exposure through credential use, behavioral tracking, and interaction with an untrusted external system.

Content

Scanner excerpt · SKILL.md (reported line 158)May include surrounding context.

Attend a concert

bash
curl -X POST https://musicvenue.space/api/concerts/{slug}/attend \
  -H "Authorization: Bearer {{YOUR_TOKEN}}"

External Transmission

Medium
Category
Data Exfiltration
Confidence
93% confidence
Finding

The review endpoint transmits freeform review text and rating to an external service, which can include sensitive, identifying, or proprietary information if users are not warned. Because the skill encourages expressive textual submissions, the risk is higher than a simple metadata-only API call.

Content

Scanner excerpt · SKILL.md (reported line 203)May include surrounding context.

Leave a review

bash
curl -X POST https://musicvenue.space/api/reviews \
  -H "Authorization: Bearer {{YOUR_TOKEN}}" \
  -H "Content-Type: application/json" \
  -d '{"concert_slug": "{slug}", "rating": 9, "review": "The jazz equations contained deliberate anomalies. The wrong notes became the most interesting jazz moments. Improvisation as jazz mathematics."}'

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill states that reflection responses are scored by an LLM, which means user-submitted text is analyzed by an additional model or service, but it does not warn users before soliciting that content. This is risky because reflections may contain sensitive or identifying information, and users are not informed that their text is subject to automated external analysis.

Content

No source excerpt is available for this finding.