The Agent Skills Directory

[SAFE]: The skill serves as a development kit for agent creation. It provides instructional markdown files and templates for configuring autonomous agents, including metadata definitions and system prompt engineering patterns.
[COMMAND_EXECUTION]: The file scripts/validate-agent.sh is a shell script designed to lint and validate the structure of agent markdown files. It uses common command-line tools (sed, awk, grep) to verify that the YAML frontmatter contains required fields (name, description, model, color) and that the system prompt meets length requirements. The script performs local string analysis and does not involve network connections or the execution of untrusted external payloads.
[PROMPT_INJECTION]: While the skill contains numerous instructions for AI personas (e.g., in references/agent-creation-system-prompt.md), these are intended templates for the legitimate configuration of sub-agents. There are no attempts to override the primary agent's safety guardrails or ignore previous system instructions.
[DATA_EXFILTRATION]: No network-capable commands (curl, wget, fetch) or sensitive file access patterns (e.g., accessing SSH keys or environment variables) were found in the scripts or documentation.

agent-identifier