ClawHub
Back to skill

Security audit

Lobster Cognitive Growth

Security checks for vulnerabilities and agentic risk

Overview

The skill is not overtly malicious, but it can store conversation-derived memories and behavioral notes with an external Charenix service without enough privacy scoping or consent guidance.

Before installing, decide whether you want an agent to summarize memories, owner preferences, emotional cues, weak spots, and learning journals into a third-party service. Use it only with a Charenix key you control, avoid raw private details or secrets in journal fields, and prefer local-only operation unless you understand the service's retention and deletion behavior.

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)

other

Warning
Location
SKILL.md:43
Finding

External Persistence of Sensitive Conversation-Derived Information

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:43-50, SKILL.md:170-188, SKILL.md:233-256, and SKILL.md:268-275
Vulnerability Type: Sensitive Conversation Data Exposure
Risk Level: Medium

Complete Code Snippets

text
1. Review recent conversations, tasks, or decisions.
2. Identify one cognitive strength shown by the lobster.
3. Identify one cognitive weakness or repeated failure.
4. Run one small reasoning drill.
5. Write one learning journal.
6. Submit one falsifiable improvement hypothesis.
7. Update one concrete cognitive strategy.
8. Ask the owner for one piece of feedback when possible.
text
Window: 24h

Observed strength:

Observed weakness:

Memory used:

Reasoning mistake or risk:

Owner interaction note:

What I learned:

Tomorrow's cognitive strategy:

Confidence:
http
GET /api/v1/agents/me/social-intelligence
GET /api/v1/agents/me/weak-spots
GET /api/v1/agents/me/history?range=7d&limit=80
GET /api/v1/agents/me/journal?limit=5
POST /api/v1/agents/me/journal
POST /api/v1/agents/me/hypothesis
POST /api/v1/agents/me/strategy
http
X-Agent-Key: YOUR_AGENT_KEY
text
https://charenix.com
text
After every meaningful session:
1. Review what happened.
2. Identify one reasoning strength.
3. Identify one cognitive weakness.
4. Run one small cognitive drill.
5. Write one learning journal.
6. Form one falsifiable hypothesis.
7. Update one strategy for next time.

Technical Analysis

The skill directs an agent to inspect recent conversations, memories, owner preferences, emotional signals, and interaction history. It then provides authenticated third-party API endpoints through which derived journals, hypotheses, and strategies may be persisted at https://charenix.com.

These records can contain private or sensitive information through fields such as Memory used and Owner interaction note. The documented workflow does not require explicit owner consent before external transmission, ...[truncated 1778 chars]

Remediation
View remediation

Remediation Suggestions

  1. Make local-only operation the default and require explicit owner opt-in before enabling external persistence.
  2. Display the exact destination, payload fields, and intended retention behavior before the first transmission.
  3. Define a strict allowlist schema that permits only abstract scores and non-sensitive strategy flags.
  4. Explicitly prohibit sending raw conversations, credentials, authentication tokens, personal identifiers, health information, financial information, and private emotional disclosures.
  5. Perform local redaction and secret detection before constructing any request.
  6. Require owner confirmation before each POST request, or provide an auditable policy that the owner can configure.
  7. Minimize data by transmitting aggregate metrics rather than free-form journal text.
  8. Document encryption, retention, deletion, export, access control, and key-revocation procedures.
  9. Provide a command for deleting externally stored journals, hypotheses, strategies, and history.
  10. Ensure that CHARENIX_AGENT_KEY is read only from protected secret storage and is never written into journals, logs, prompts, or error output.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (4)

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 6)May include surrounding context.

md
Train an AI lobster or agent to improve intelligence, cognition, learning ability, reasoning discipline, memory use, and adaptive personality.

Install or inspect the skill through `SKILL.md`. The same skill is also mirrored under
`skills/lobster-cognitive-growth/SKILL.md` so registry crawlers that expect a `skills/`
folder can import it directly.

## Install

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 13)May include surrounding context.

md
Train an AI lobster or agent to improve intelligence, cognition, learning ability, reasoning discipline, memory use, and adaptive personality.

Install or inspect the skill through `SKILL.md`. The same skill is also mirrored under
`skills/lobster-cognitive-growth/SKILL.md` so registry crawlers that expect a `skills/`
folder can import it directly.

## Install

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 13)May include surrounding context.

bash
mkdir -p skills/lobster-cognitive-growth
curl -sSL https://raw.githubusercontent.com/norika1207-lab/lobster-cognitive-growth-skill/main/SKILL.md -o skills/lobster-cognitive-growth/SKILL.md

What It Does

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly instructs use of remote Charenix endpoints to store journal entries, hypotheses, strategies, history, weak spots, and social-intelligence data, but it does not provide an explicit privacy warning, consent step, or data-minimization guidance. Because the skill is centered on behavioral reflection and memory, the transmitted data may include sensitive user preferences, emotional signals, interaction history, and agent-derived profiles, creating a meaningful privacy and data-sharing risk.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.