gh-actions-validator

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (HIGH): Vulnerable to Indirect Prompt Injection (Category 8). The skill is designed to ingest untrusted data from .github/workflows/ files. Mandatory Evidence: 1. Ingestion points: File reads from the .github/workflows/ directory. 2. Boundary markers: Absent; no instructions provide delimiters for the untrusted content. 3. Capability inventory: The skill utilizes Bash(gcloud:*) and filesystem Write/Edit tools. 4. Sanitization: Absent. An attacker could place malicious instructions inside a workflow's YAML comments or metadata to trick the agent into executing unauthorized gcloud commands or exfiltrating sensitive cloud configurations.
  • [COMMAND_EXECUTION] (MEDIUM): The skill requires Bash(gcloud:*) access and administrative gcloud permissions. While the stated goal is hardening, the broad capability to interact with Google Cloud resources poses a significant risk if the agent's logic is subverted via the aforementioned prompt injection vulnerability.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 12:42 PM