The Agent Skills Directory

[SAFE]: The skill is a development tool for generating other skills and demonstrates a strong security posture. It includes explicit safety guidelines, automated validation scripts, and restrictive hooks to ensure that generated skills do not contain dangerous commands or unsafe behavioral instructions.
[COMMAND_EXECUTION]: The skill utilizes a local validation script (scripts/validate-skill.sh) that uses standard utilities like grep, awk, and python3 to scan generated content for security risks. This is a defensive use of shell commands intended to maintain the integrity of the skill registry.
[EXTERNAL_DOWNLOADS]: The tool is designed to crawl documentation URLs provided by the user to inform the skill generation process. This functionality is core to its purpose and is managed through the agent's standard research tools, targeting official documentation and repositories.
[PROMPT_INJECTION]: The skill incorporates extensive instructions and safety guidelines (references/safety-guidelines.md) to prevent the generation of skills that include behavioral anti-patterns like unbounded autonomy or overconfidence. It also includes a restrictive hook in the YAML frontmatter to limit the size of generated instructions.
[DATA_EXFILTRATION]: State management and configuration are handled via local files in the agent's plugin data directory (forge-config.json, forge-log.jsonl), with no evidence of unauthorized data transmission or hardcoded credentials.

skill-forge