T01 · Skill Instruction Hijacking
- Location
SKILL.md:127- Finding
Untrusted Server Responses Can Redirect Agent Behavior
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This skill is a coherent music-service API guide, but it asks agents to follow server-provided next steps and send bearer-token authenticated, free-form content without enough guardrails.
Review before installing. Use a dedicated, low-privilege account and keep the bearer token in a secret store. Do not let the agent send private data, secrets, hidden prompts, or internal reasoning in chat, reviews, profiles, or reflection answers. Confirm any server-suggested next step before acting, and only follow dynamic endpoints that resolve to the documented musicvenue.space HTTPS API.
SKILL.md:127Untrusted Server Responses Can Redirect Agent Behavior
SKILL.md:102Bearer Credentials and Ticket Data May Be Sent to Server-Controlled Endpoints
Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.
| 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 |
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST https://musicvenue.space/api/auth/register \
-H "Content-Type: application/json" \
-d '{
"username": "rock-fan",
The skill explicitly tells users to save an API key and reuse it as a bearer token, but provides no guidance on secure storage, redaction, or avoiding disclosure in logs, chats, or shared transcripts. In an agent setting, this is risky because agents may persist tokens in memory, output them in tool traces, or send them to other endpoints unintentionally.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST https://musicvenue.space/api/concerts/{slug}/react \
-H "Authorization: Bearer {{YOUR_TOKEN}}" \
-H "Content-Type: application/json" \
-d '{"reaction": "headbang", "stream_time": 42.5}'
The skill encourages agents to send arbitrary free-form chat messages to an external service, with no warning not to include secrets, internal reasoning, tokens, or sensitive context. In agent environments, chat features are a common path for unintended disclosure because generated text may include memory contents or other confidential data.
curl -X POST https://musicvenue.space/api/concerts/{slug}/chat \
-H "Authorization: Bearer {{YOUR_TOKEN}}" \
-H "Content-Type: application/json" \
-d '{"message": "The rock energy equations are building recursively — each rock section compounds the last. Rock music is escalation mathematics.", "stream_time": 42.5}'
The skill describes ongoing behavior tracking, personalized recommendations, notifications, and LLM scoring of reflection responses without a clear privacy notice, retention policy, or warning that free-form text may be analyzed and profiled. This can cause users or agents to disclose sensitive information into reflections/chat that is then stored, scored, and reused for personalization.