ClawHub
Back to skill

Security audit

golang-continuous-integration

Security checks across malware telemetry and agentic risk

Overview

This is a legitimate Go CI/CD skill, but its AI review workflow ships with broad automatic PR-commenting and log-reading authority that should be reviewed before installation.

Review before installing on any important repository. Consider scoping the Claude AI review workflow to trusted contributors, protected branches, labels, or manual dispatch; move permissions to the jobs that need them; remove `id-token: write` unless required; pin or narrow the remote skill installation; and remove the hidden setup comment before copying Copilot instructions. Keep branch protection and required checks enabled before using the auto-merge, release, or Docker publishing templates.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The workflow grants repository-wide `pull-requests: write` and `actions: read` permissions, then triggers on untrusted pull-request activity and review events. That lets an LLM-driven action post, edit, or resolve comments across PRs and inspect workflow logs that may contain sensitive operational data, increasing blast radius if the agent is misprompted, compromised, or abused through adversarial PR content.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The workflow activates on multiple PR and review-related events without strong scope constraints, causing an LLM-powered reviewer to run on broad untrusted input surfaces. In this skill context, that is more dangerous because the action is explicitly designed to read attacker-controlled PR content and then use write-capable GitHub tools, increasing the chance of abuse, noisy automation, privilege misuse, or unintended data exposure through comments and log inspection.

Hidden Instructions

High
Category
Prompt Injection
Content
<!-- Prerequisites:
  The skills CLI (listed in the frontmatter install block) can be used to copy skills locally:
    npx skills add https://github.com/samber/cc-skills-golang --agent github-copilot --skill '*' -y --copy
    ln -s .agents .copilot
Confidence
85% confidence
Finding
The HTML comment contains hidden operational instructions telling the agent/user how to fetch and install skills from a remote GitHub repository, including copying all skills via a wildcard. Hidden instructions are dangerous because downstream agents may parse or obey commented text despite humans treating it as non-operative, enabling prompt injection and unreviewed retrieval of external content into the trust boundary.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.