Back to skill

Security audit

create-all-cg-save

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners found save-editing actions that could overwrite progress or alter switches beyond CG and gallery unlocks.

Consider the risk of save damage or unintended game-state changes before installing. The scanners report that rollback may write during a dry run and that `--all` may affect switches beyond unlocks; the scan was partial because some local references were unresolved.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The save-editing instruction omits a warning about damaging existing save data.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The decision guide explicitly instructs the agent to modify JSON save files, rather than merely listing a tool or format. It does not warn that syntactically valid changes can still corrupt game state or make a save unusable, and provides no backup precaution.

Content

No source excerpt is available for this finding.

The globally advertised --dry-run option does not prevent rollback from overwriting the save.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The CLI accepts --dry-run with rollback, but dispatches to cmd_rollback without passing or checking that flag. Rollback consequently overwrites the current save even when the user requested a non-writing preview, potentially destroying progress made since the backup; backup integrity checks do not prevent this loss.

Content

No source excerpt is available for this finding.

The --all option generates patches for every game switch, not just switches controlling CG or gallery unlocks.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The --all path creates an unlock plan for every allocated switch without distinguishing gallery flags from story, branching, or failure-state flags. In a CG-unlocking workflow, applying this plan can produce a structurally valid but logically broken save; explicit opt-in and the separate application step reduce accidental exposure but do not validate the changes' semantic scope.

Content

No source excerpt is available for this finding.