The Agent Skills Directory

[SAFE]: A comprehensive analysis was performed across all threat categories, including prompt injection, data exfiltration, and remote code execution. No malicious patterns or vulnerabilities were detected.
[COMMAND_EXECUTION]: The skill allows users and agents to associate shell commands with validation gates (e.g., 'npm test'). However, technical analysis of gate.py confirms that these strings are treated strictly as metadata; the script does not contain any calls to os.system, subprocess, eval, or other execution sinks. The documentation in SKILL.md explicitly clarifies that recording a gate does not run the command.
[INDIRECT_PROMPT_INJECTION]: The skill processes external strings such as gate names and notes which are later presented to the agent. While this represents a data ingestion surface, the skill is considered safe because it does not possess any sensitive capabilities (such as network access or file system writes outside its state directory) that could be exploited by malicious content within those strings.

quality-gate-orchestrator