Security audit
Security checks for vulnerabilities and agentic risk
The skill is coherent for multi-agent task orchestration, but it gives broad autonomous agent, file-copying, command, and retention authority without enough user-control safeguards.
Install only if you intentionally want a skill that coordinates autonomous sub-agents. Before using it, choose a dedicated workspace, avoid giving it sensitive files unless necessary, confirm which files may be copied, require approval before Bash or write-heavy subtasks, and decide whether agent workspaces should be archived or deleted after completion.
Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.
# Task: {TASK_NAME}
## Objective
{Clear statement of what needs to be accomplished}
## Context
{Background information relevant to the task}
Suspicious Unicode normalization or mixed-script content
The mandatory trigger list includes very broad phrases such as 'orchestrate', 'decompose task', and 'delegate tasks', which can cause the skill to activate in routine conversations where the user did not intend autonomous multi-agent behavior. In this skill, unintended activation is more dangerous because activation can lead to spawning sub-agents, creating workspaces, and copying files, expanding the action surface beyond a normal single-agent skill.
This section instructs the operator to create sub-agent workspaces and generate per-agent SKILL.md files, but it does not clearly warn that the process will create directories and persistent files on disk. In the context of an orchestration skill, that omission is risky because users may unknowingly authorize broad filesystem changes and the creation of autonomous agent instructions that persist beyond the immediate task.
The dispatch phase tells the orchestrator to copy required input files into each agent inbox without clearly warning the user that their files may be duplicated across multiple agent workspaces. This increases data exposure risk, especially if sensitive files are propagated to several sub-agents or archived later, multiplying the number of locations where confidential content resides.
The dissolution step mentions archiving workspaces and cleaning up temporary files, but it does not clearly warn users that agent artifacts may be retained or moved into archives instead of being deleted. In a multi-agent system that may handle copied inputs, generated instructions, and outputs, unclear archival behavior can lead to unintended long-term retention of sensitive material.
The Code Agent template explicitly authorizes file writes and Bash command execution, including running tests and fixing issues, but provides no safety constraints, approval requirements, or user-visible warning about the side effects of those actions. In an agent-orchestration skill that dynamically spawns sub-agents, this increases the chance that generated agents will perform unintended filesystem changes or execute risky commands based only on task input.
No suspicious patterns detected.