Back to skill

Security audit

Build Pixel Pet Sprites

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent sprite-building workflow, but its bundled scripts trust manifest and config paths in ways that can read images or write generated files outside the intended folders.

Use this only with sprite configs and packages you trust, preferably inside a project-limited or disposable workspace. Be cautious with third-party sprite-manifest.json or package-config.json files until the scripts validate plain filenames, reject traversal, and verify all reads and writes stay under the chosen package directories.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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 (3)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/build_sprite_package.py:43
Finding

Configuration-Controlled Path Traversal Enables Arbitrary File Writes

Content
View full analysis
Remediation
View remediation
Path: root = root.resolve() if relative.is_absolute(): raise ValueError("absolute output paths are not allowed") candidate = (root / relative).resolve() if not candidate.is_relative_to(root): raise ValueError("output path escapes the package directory") return candidate ``` 4. Apply containment validation to atlas, frame, strip, and preview destinations. 5. Reject symlinks in writable package path components where packages may be processed in attacker-controlled directories. 6. Add tests covering absolute paths, nested traversal, mixed separators, and symlink escapes. ]]>

T09 · Insecure Skill Coding Practices

Error
Location
scripts/align_sprite_frames.py:235
Finding

Untrusted Sprite Manifest Enables Arbitrary Image Reads and Filesystem Writes

Content
View full analysis
Remediation
View remediation
Path: root = root.resolve() relative = Path(value) if relative.is_absolute(): raise ValueError("absolute manifest paths are not allowed") candidate = (root / relative).resolve(strict=True) if not candidate.is_relative_to(root): raise ValueError("manifest path escapes the source package") return candidate ``` 5. Resolve every output destination and verify that it remains beneath the resolved output root before creating directories or writing. 6. Account for symbolic links by checking fully resolved paths. Consider rejecting symlinked source or output path components when processing packages from untrusted users. 7. Validate manifest structure and types with a schema before processing images, including dimensions, rows, columns, frame counts, action names, and path fields. 8. Add negative tests for absolute paths, traversal sequences, symlink escapes, and platform-specific separators. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/validate_sprite_package.py:39
Finding

Validator Trusts Manifest Paths and Can Read Images Outside the Package

Content
View full analysis
Remediation
View remediation
Path: package = package.resolve() relative = Path(value) if relative.is_absolute(): raise ValueError("absolute manifest paths are not allowed") candidate = (package / relative).resolve(strict=True) if not candidate.is_relative_to(package): raise ValueError("manifest path escapes the package") return candidate ``` 2. Reject `..`, empty path components, unexpected separators, and unsupported file extensions before resolving paths. 3. Check resolved paths to prevent symbolic-link escapes. 4. Apply the same validation consistently to atlas, strip, and frame references. 5. Add image resource limits, including maximum dimensions and pixel counts, before full conversion and analysis. 6. Validate the manifest against a strict schema and return a controlled validation error instead of opening invalid references. 7. Add tests for absolute paths, traversal paths, symlink escapes, nonexistent paths, and oversized images. ]]>
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill instructs the agent to read references and run local scripts that create, align, validate, and package files, which implies file read/write capabilities. Because the skill does not declare any tool scope or allowed-tools boundary, an agent may execute these filesystem operations without an explicit least-privilege restriction, increasing the chance of unintended file access or overwriting in the broader workspace.

Content

No source excerpt is available for this finding.