The Agent Skills Directory

[SAFE]: The skill is purely instructional and provides a disciplined workflow for prompt optimization based on industry standards and research.
[EXTERNAL_DOWNLOADS]: The skill references official documentation and research papers from trusted organizations such as OpenAI, Anthropic, Google, and arXiv. These references are used for provenance and technical guidance rather than runtime code execution.
[DATA_EXFILTRATION]: No network operations, hardcoded credentials, or access to sensitive local file paths (such as .ssh, .aws, or .env) were detected.
[PROMPT_INJECTION]: The instructions do not attempt to override agent safety protocols, bypass restrictions, or extract system prompts. Instead, it provides frameworks to improve prompt reliability and instruction-following.
[INDIRECT_PROMPT_INJECTION]: The skill acknowledges an attack surface by processing user-supplied prompts. It mitigates this risk by explicitly instructing the agent to use clear boundary markers (XML tags or Markdown headings) to separate durable instructions from potentially untrusted task-local data.

prompt-optimizer