Security audit

git-workspace-review

Security checks for vulnerabilities and agentic risk

Overview

The skill is a Git review helper, but it claims to be read-only while instructing agents to run repository-defined commands and change staging or file contents.

Review this skill carefully before installing. It is not just observational: an agent following it may run Makefile targets, stage or unstage files, and format code. Use it only in repositories you trust, and require explicit confirmation before any command that changes files, changes the Git index, or executes project-defined targets.

A.I.G

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 (2)

T09 · Insecure Skill Coding Practices

Error

Location: SKILL.md:50
Finding: Untrusted Repository Commands Executed Through Make Targets
Content: View full analysis
Remediation: View remediation

T09 · Insecure Skill Coding Practices

Warning

Location: SKILL.md:46
Finding: Read-Only Review Workflow Directs Unauthorized Workspace and Index Mutations
Content: View full analysis
Remediation: View remediation

SkillSpector

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (8)

Intent-Code Divergence

High

Category: Not specified by scanner
Confidence: 97% confidence
Finding: The top-level description claims the skill is a read-only preflight, yet later instructions include commands that can rewrite files or alter staged state, such as formatting, lint-driven fixes, and stage/unstage guidance. This kind of semantic mismatch is dangerous because safety policy, user expectations, and automation routing may all rely on the manifest description when deciding whether the skill is safe to invoke.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

High

Category: Not specified by scanner
Confidence: 97% confidence
Finding: The skill markets itself as a read-only preflight, but these steps tell the agent or user to change repository state by staging/unstaging files and by fixing issues immediately. That mismatch can cause unexpected write actions in a context where downstream systems or users may have granted trust based on the read-only description, increasing the risk of unintended repository modifications.
Content: No source excerpt is available for this finding.

Shadow Command Trigger

Medium

Category: Trigger Abuse
Confidence: 90% confidence
Finding: Using 'status' as a trigger can shadow or collide with a built-in command of the same name, causing this skill to activate when a user intended a standard tool action. In a skill that is presented as read-only but includes mutating guidance, such command ambiguity raises the risk of unintended workflow execution and confusing or unsafe behavior.
Content: No source excerpt is available for this finding.

Shadow Command Trigger

Medium

Category: Trigger Abuse
Confidence: 90% confidence
Finding: Using 'diff' as a trigger can shadow or conflict with a built-in 'diff' command, making accidental invocation more likely during normal development workflows. Since the skill contains broader workflow instructions than a simple diff display, this collision can lead to unexpected behavior and unsafe task routing.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 90% confidence
Finding: The triggers include very generic terms like 'git', 'status', and 'diff', which are likely to match ordinary conversations and invoke the skill unexpectedly. Because the skill contains instructions that can lead to repository changes, overbroad triggering increases the chance of accidental activation in unrelated or only partially related contexts.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium

Category: Not specified by scanner
Confidence: 85% confidence
Finding: The skill context describes a read-only preflight used before commits or PRs to verify workspace state and staged changes. git log --oneline -10 and git show <commit-hash> add commit-history and arbitrary commit inspection capabilities, which are not necessary to confirm current workspace state or the staged set.
Content: No source excerpt is available for this finding.

Missing User Warnings

Low

Category: Not specified by scanner
Confidence: 86% confidence
Finding: The instruction to stage or unstage files appears without a prominent warning that this modifies repository state. In isolation this is a modest issue, but in the context of a purported read-only preflight it materially increases the chance that a user or agent performs unintended state changes.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

Low

Category: Not specified by scanner
Confidence: 78% confidence
Finding: The stated purpose is to confirm the workspace state and staged changes as a preflight. Comparing origin/main...HEAD reviews the entire branch delta against main, which goes beyond checking the current staged set and workspace cleanliness.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.