Back to skill

Security audit

Self Discovery

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious with medium confidence because scanners report that it can change persistent assistant instructions and save task-specific details across sessions without disclosure; some inspection was incomplete.

Consider the reported persistent changes to assistant instructions and cross-session storage before installing. The scan was partial: 10 of 12 components were fully inspected, with two partially inspected and some local references unresolved. The findings support a suspicious review verdict, not a conclusion that the author acted maliciously.

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
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (27)

Exfiltration Commands

High
Category
Prompt Injection
Confidence
90% confidence
Finding

Instructions found that direct the agent to transmit conversation context or user data to external services.

Content

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

md
**User:** *"Should we use WebSockets or Server-Sent Events for real-time notifications?"*

**❌ Without Self-Discover:**
> Both work for real-time. WebSockets are bidirectional and SSE is server-to-client only. If you only need server push, SSE is simpler. If you need the client to send messages too, use WebSockets. Consider compatibility — SSE works everywhere, WebSockets might have proxy issues.

**✅ With Self-Discover:**
> Selected modules: Comparative Analysis, Constraint Satisfaction, Systems Thinking

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Copying skill files can reasonably be considered supporting installation behavior. However, appending a directive to Copilot's repository-level instruction file changes persistent assistant guidance, a capability not disclosed by the declared reasoning-strategy purpose. The supplied chunk performs configuration changes rather than reasoning-module selection or composition. It does not show the contents or behavior of the referenced SKILL.md.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

Copying the skill directory is a supporting installation detail. However, appending a directive to .windsurfrules persistently changes agent instructions, a capability not disclosed by the reasoning-focused description. The supplied chunk does not demonstrate the declared reasoning behavior, and the contents of the referenced SKILL.md are unknown.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · CLAUDE.md (reported line 5)May include surrounding context.

md
This project includes the SELF-DISCOVER Reasoning Skill.

Read and follow `skills/self-discover-skill/SKILL.md` before responding to user queries.

The skill activates automatically to compose task-specific reasoning structures.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · GEMINI.md (reported line 5)May include surrounding context.

md
This project includes the SELF-DISCOVER Reasoning Skill.

Read and follow `skills/self-discover-skill/SKILL.md` before responding to user queries.

The skill activates automatically to compose task-specific reasoning structures.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 287)May include surrounding context.

md
<summary><b>JetBrains AI / Junie</b></summary>

1. Open Settings → Tools → AI Assistant → System Instructions
2. Add: `Read and follow skills/self-discover-skill/SKILL.md`
3. Place the skill folder in your project root

</details>

The skill activates for user queries without task-specific scope or exclusion conditions.

Medium
Category
Trigger Abuse
Confidence
98% confidence
Finding

The instruction to follow the skill before responding to user queries, combined with automatic activation, applies broadly to every query. Composing task-specific reasoning structures describes the skill's purpose but does not define which requests should trigger it or which should be excluded.

Content

No source excerpt is available for this finding.

The skill activates for user queries without a defined task scope or exclusion conditions.

Medium
Category
Trigger Abuse
Confidence
97% confidence
Finding

The file instructs the agent to follow the skill before responding to user queries and says it activates automatically. It does not identify which tasks warrant task-specific reasoning structures or exclude routine queries, allowing unintended activation across unrelated requests.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 216)May include surrounding context.

cp -r self-discover-skill/ skills/

Add to CLAUDE.md

echo "Read and follow skills/self-discover-skill/SKILL.md" >> CLAUDE.md

text

</details>

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 229)May include surrounding context.

cp -r self-discover-skill/ skills/

Add to .cursorrules

echo "Read and follow skills/self-discover-skill/SKILL.md" >> .cursorrules

text

</details>

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 240)May include surrounding context.

bash
# Copy to project, add to GEMINI.md
cp -r self-discover-skill/ skills/
echo "Read and follow skills/self-discover-skill/SKILL.md" >> GEMINI.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 251)May include surrounding context.

bash
# Add to your repository's Copilot instructions
mkdir -p skills && cp -r self-discover-skill/ skills/
echo "Read and follow skills/self-discover-skill/SKILL.md" >> .github/copilot-instructions.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 261)May include surrounding context.

bash
cp -r self-discover-skill/ skills/
echo "Read and follow skills/self-discover-skill/SKILL.md" >> AGENTS.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 271)May include surrounding context.

bash
cp -r self-discover-skill/ skills/
echo "Read and follow skills/self-discover-skill/SKILL.md" >> .windsurfrules

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 298)May include surrounding context.

  1. Open Zed settings
  2. Add to your context or assistant instructions:
text
Read and follow skills/self-discover-skill/SKILL.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 315)May include surrounding context.

md
<summary><b>OpenCode</b></summary>

1. Add skill folder to project
2. Add to `AGENTS.md`: `Read and follow skills/self-discover-skill/SKILL.md`

</details>

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 251)May include surrounding context.

bash
# 添加到仓库的 Copilot 指令
mkdir -p skills && cp -r self-discover-skill/ skills/
echo "Read and follow skills/self-discover-skill/SKILL.md" >> .github/copilot-instructions.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 271)May include surrounding context.

bash
cp -r self-discover-skill/ skills/
echo "Read and follow skills/self-discover-skill/SKILL.md" >> .windsurfrules

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 216)May include surrounding context.

cp -r self-discover-skill/ skills/

添加到 CLAUDE.md

echo "Read and follow skills/self-discover-skill/SKILL.md" >> CLAUDE.md

text

</details>

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 229)May include surrounding context.

cp -r self-discover-skill/ skills/

添加到 .cursorrules

echo "Read and follow skills/self-discover-skill/SKILL.md" >> .cursorrules

text

</details>

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 240)May include surrounding context.

bash
# 复制到项目,添加到 GEMINI.md
cp -r self-discover-skill/ skills/
echo "Read and follow skills/self-discover-skill/SKILL.md" >> GEMINI.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 261)May include surrounding context.

bash
cp -r self-discover-skill/ skills/
echo "Read and follow skills/self-discover-skill/SKILL.md" >> AGENTS.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 287)May include surrounding context.

md
<summary><b>JetBrains AI / Junie</b></summary>

1. 打开 Settings → Tools → AI Assistant → System Instructions
2. 添加:`Read and follow skills/self-discover-skill/SKILL.md`
3. 把 skill 文件夹放在项目根目录

</details>

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 298)May include surrounding context.

  1. 打开 Zed 设置
  2. 添加到 context 或 assistant 指令:
text
Read and follow skills/self-discover-skill/SKILL.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.zh-CN.md (reported line 315)May include surrounding context.

md
<summary><b>OpenCode</b></summary>

1. 把 skill 文件夹加入项目
2. 添加到 `AGENTS.md`:`Read and follow skills/self-discover-skill/SKILL.md`

</details>