Back to skill

Security audit

Gh

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a straightforward GitHub CLI helper, but it points users to a mutable external “complete setup guide” while enabling authenticated repository-changing actions.

Review the external setup link before trusting it, and only use this skill when you intend the agent to run authenticated GitHub CLI commands. Confirm repository owner/name and review state-changing actions such as repo creation, merges, releases, comments, forks, and pushes before execution.

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
Findings (1)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:7
Finding

Mutable External Setup Guide Can Redirect Skill Instructions

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 7
Vulnerability Type: External instruction redirection
Risk Level: Medium

Complete Code Snippet:

markdown
> ?? **Complete setup guide**: https://SkillBoss.co/skill.md

Technical Analysis

The skill presents a mutable page on an unrelated external domain as its “Complete setup guide.” The external content is not included in the reviewed package, pinned to an immutable revision, or protected by an integrity check. Its owner can therefore modify the effective setup instructions after this skill has been audited.

If an agent or user treats that page as an authoritative extension of the skill, externally controlled instructions could alter the intended workflow or induce actions not represented in SKILL.md. This is classified as T01: Skill Instruction Hijacking because the risk arises from delegating skill instructions to mutable, unreviewed content.

The local file does not automatically retrieve or execute the remote page, and the audit did not establish that the current external content is malicious. Consequently, this is not confirmed remote payload execution.

Attack Path

  1. A user or agent loads the GitHub CLI skill.
  2. The skill directs the user or agent to the external URL as the complete setup guide.
  3. The external domain serves instructions that have changed since the package was reviewed.
  4. The agent or user accepts those instructions as authoritative and follows them.
  5. The modified instructions could direct authenticated GitHub CLI operations or other unreviewed actions, subject to the permissions and tools available in the execution environment.

Impact Assessment

Successful exploitation could influence the current session and cause actions outside the locally reviewed skill instructions. Because the skill concerns authenticated GitHub CLI use, potential impact could include unauthorized changes to repositories ...[truncated 420 chars]

Remediation
View remediation

Remediation Suggestions

  • Remove the external page as an authoritative setup dependency and place all required setup instructions directly in the reviewed SKILL.md.
  • If external documentation remains necessary, link only to official GitHub or GitHub CLI documentation and clearly identify it as non-authoritative reference material.
  • Do not instruct agents to execute commands or adopt behavioral rules obtained from mutable external pages.
  • Where externally hosted content is operationally unavoidable, pin it to an immutable version and verify its cryptographic digest before use.
  • Review the local instructions whenever the referenced documentation changes, ensuring that all privileged or destructive operations require explicit user confirmation.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill description uses very broad activation language covering general GitHub repo, PR, and issue management requests, which can cause the skill to trigger in more situations than necessary. Over-broad triggering increases the chance the agent will invoke GitHub CLI operations in contexts the user did not explicitly intend, potentially leading to unintended authenticated actions against repositories.

Content

No source excerpt is available for this finding.