promql-validator

Warn

Audited by Gen Agent Trust Hub on Mar 28, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The workflow in SKILL.md instructs the agent to execute shell commands using user-provided input: python3 devops-skills-plugin/skills/promql-validator/scripts/validate_syntax.py "<query>". An attacker could provide a malicious query containing shell metacharacters (e.g., "; touch /tmp/exploited #) to execute arbitrary commands on the host system.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user data (the PromQL query) without providing the agent with clear instructions for sanitization or escaping.
  • Ingestion points: User-supplied PromQL queries are passed to shell commands in SKILL.md.
  • Boundary markers: The instructions suggest using double quotes ("<query>") but do not include explicit sanitization requirements or warnings to ignore embedded instructions.
  • Capability inventory: The agent has the capability to execute shell commands (python3, git) and read/write files in the local repository.
  • Sanitization: There is no evidence of input validation or escaping logic in the instructions or the supporting Python scripts.
  • [COMMAND_EXECUTION]: The skill requires the agent to establish its working directory by running git rev-parse --show-toplevel. While common for development tools, this involves automated shell execution in the local environment.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 28, 2026, 06:18 PM