Back to skill

Security audit

Code Review Auditor

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed code-review helper that writes local review reports, with no evidence of hidden network access, credential use, background persistence, or destructive behavior.

Install only if you are comfortable with each run creating a local review/<timestamp>/ directory in the reviewed repository and recording basic repo metadata. Consider ignoring or cleaning those report files if you do not want them committed, and use fix/refactor modes only when you intend to approve or request code changes.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (10)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description emphasizes analysis without changing code during analysis, yet the body mandates creating directories and writing multiple artifacts into the target project. This mismatch can mislead users and higher-level safety systems about side effects, causing unapproved workspace modification and trust-boundary violations in environments where review actions are expected to be read-only.

Content

No source excerpt is available for this finding.

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 · README.md (reported line 64)May include surrounding context.

text

```text
Use $code-review-auditor to create a refactoring plan for the hotspots, but do not change code yet.

Installation

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill appears to require shell and file-write capabilities but does not declare any explicit tool scope such as permissions or allowed-tools. That creates an authority mismatch: a review-oriented skill can perform workspace-modifying actions without transparent least-privilege boundaries, making accidental or unauthorized file creation more likely.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/create_review_run.py (reported line 31)May include surrounding context.

python
def git_value(root: Path, *args: str) -> str:
    try:
        result = subprocess.run(
            ["git", *args],
            cwd=root,
            check=True,

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · rules/categories.md (reported line 92)May include surrounding context.

md
## Performance

- Look for missing or unused indexes, non-sargable predicates, full scans, sort/hash pressure, N+1 access, unbounded queries, and excessive joins.
- Check pagination strategy, batch size, connection pool settings, lock duration, and migration impact on large tables.

## Architecture

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · stacks/database.md (reported line 18)May include surrounding context.

md
## Performance

- Look for missing or unused indexes, non-sargable predicates, full scans, sort/hash pressure, N+1 access, unbounded queries, and excessive joins.
- Check pagination strategy, batch size, connection pool settings, lock duration, and migration impact on large tables.

## Architecture

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill metadata says analysis should occur without changing code, but the documented fix and refactor modes authorize creating a plan file and potentially implementing changes after approval or when already requested. This creates a capability mismatch that can cause downstream agents or users to invoke code-changing behavior under a skill that appears read-only, weakening trust boundaries and increasing the chance of unintended repository modification.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

This markdown file states that every run creates a timestamped review directory with multiple output files in the target project. Although the behavior is documented, it does not explicitly warn users that invoking the skill will write new files into their repository, which is a user-impacting filesystem change.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill requires writing review artifacts into the project tree at the end of every execution without an explicit user-facing warning that the workspace will be modified. Even if the writes are intended as benign reports, silent modification of the repository can pollute working trees, interfere with automation, or surprise users who requested analysis only.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

This code creates a new review directory and writes multiple markdown and JSON files, including repository path, branch, commit, and dirty-state metadata. While the behavior is visible in code and implied by the module docstring, there is no confirmation prompt or explicit user-facing warning at runtime before persisting these artifacts into the target project.

Content

No source excerpt is available for this finding.