The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bash script located at scripts/validate_skill.sh to perform static analysis on SKILL.md files. This script uses standard utilities such as awk, grep, and sed to enforce structural constraints and portability rules without introducing harmful side effects.\n- [EXTERNAL_DOWNLOADS]: The reference documentation contains numerous links to official developer resources and prompt engineering guides from trusted organizations and well-known services, including OpenAI, Anthropic, Microsoft, and Vercel. These links are provided solely for developer reference and do not involve piped remote execution.\n- [INDIRECT_PROMPT_INJECTION]: As a meta-tool for building and auditing other skills, the agent processes external instructional content. The skill effectively manages this attack surface by providing a dedicated audit framework (frontier-model-audit.md) and a validation script that identifies common security hazards such as absolute paths or incomplete instructions.\n- [DYNAMIC_EXECUTION]: The scripts/smoke_test.sh script dynamically generates temporary skill structures and files within the system's temporary directory to verify the validator's logic. This follows standard software testing practices and is used to ensure the reliability of the diagnostic tools.\n- [CREDENTIALS_UNSAFE]: The skill provides explicit security guidelines in references/local-skill-layout.md instructing authors never to copy secrets, tokens, or local credentials into repository folders, promoting industry-standard secret management.

skill-builder