Back to skill

Security audit

Fox Cellcog

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent CellCog cloud-service integration, but users should understand that tagged local files can be uploaded and generated files can be written locally.

Before installing, treat CellCog as a third-party cloud processor: only tag files you are comfortable uploading, avoid secrets or regulated data unless approved, review generated files before opening them, and consider installing the unpinned SDK in an isolated environment.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:8
Finding

Unpinned Third-Party Python Dependency

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 8–10 and 72–76
Vulnerability Type: Unpinned third-party dependency
Risk Level: Medium

Vulnerable Code

yaml
install:
  pip: cellcog
python
from cellcog import CellCogClient
bash
pip install cellcog

Technical Analysis

The skill directs users to install the cellcog package without an exact version constraint, cryptographic hash, signature, or verifiable source repository. Consequently, the installed artifact depends on whichever release the package index resolves at installation time.

The dependency's implementation is not included in this project, so its installation hooks and runtime behavior could not be audited. Python packages may execute package-controlled build or installation logic, while the installed SDK will run with the invoking user's privileges. A compromised or unexpectedly modified future release could therefore affect users without any change to this reviewed skill.

Attack Path

  1. An attacker compromises the package publisher, distribution account, package-index channel, or a future release process.
  2. The attacker publishes a malicious or backdoored version under the same package name.
  3. A user or agent follows the documented pip install cellcog instruction.
  4. Package resolution retrieves the attacker-controlled release because no version or hash is enforced.
  5. Malicious logic executes during installation, import, or SDK use with the invoking user's permissions.
  6. The payload may access data available to that process, including the configured API key and files explicitly supplied to the SDK.

Impact Assessment

Successful exploitation could permit arbitrary code execution within the installing user's security context. The reachable scope may include the CELLCOG_API_KEY, user-readable files, local CellCog state, generated artifacts, and network resources available to the process.

No evidence shows that the currently publis ...[truncated 127 chars]

Remediation
View remediation

Remediation Suggestions

  • Pin the dependency to an exact, reviewed version, such as cellcog==X.Y.Z.
  • Distribute and enforce cryptographic hashes through a lock file or pip install --require-hashes.
  • Identify the authoritative source repository and document how package provenance can be verified.
  • Use signed releases or trusted-publisher attestations where supported.
  • Review the package source, build configuration, installation hooks, and transitive dependencies before approving a version.
  • Install and run the SDK in an isolated, least-privilege virtual environment or container.
  • Restrict filesystem and network access to what the intended cloud-processing workflow requires.
  • Rotate the API key and inspect local state if dependency compromise is suspected.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 89)May include surrounding context.

export CELLCOG_API_KEY="sk_..."

text

Get API key from: https://cellcog.ai/profile?tab=api-keys

Check configuration:
```python

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 351)May include surrounding context.

python
prompt = """
Analyze this sales data and create a report:
<SHOW_FILE>/path/to/sales.csv</SHOW_FILE>
"""

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill explicitly states that files referenced with <SHOW_FILE> are transmitted to CellCog's cloud service, but the privacy warning appears later in the document rather than upfront where file-upload behavior is introduced. In an agent setting, this increases the risk of sensitive local files being sent to a third party without sufficiently prominent disclosure or consent boundaries.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The skill says generated files are automatically downloaded to the local machine, which is a filesystem side effect that can write arbitrary model-generated artifacts into user-accessible locations. Without a clear warning near the first mention of this behavior, users or calling agents may not anticipate local writes, storage consumption, or accidental placement of untrusted files.

Content

No source excerpt is available for this finding.