The Agent Skills Directory

[DATA_EXPOSURE]: The skill includes a 'Project Detection' phase that reads standard project configuration files (e.g., package.json, go.mod, requirements.txt, GEMINI.md) to identify the technical stack. This access is localized and intended to provide contextual optimization advice without evidence of external data transmission.
[COMMAND_EXECUTION]: While the skill frequently references and recommends platform-specific commands (such as /egc-plan, /egc-tdd, and /egc-verify), it contains explicit instructions to remain in an 'advisory role only'. It is specifically directed not to write code, create files, or execute any commands itself.
[PROMPT_INJECTION]: The skill incorporates internal logic to prevent scope creep, instructing the agent to ignore user prompts that demand direct execution (e.g., 'just do it') and instead redirect them to the optimization workflow. This serves as a behavioral guardrail rather than a malicious injection.
[NO_CODE]: The skill consists entirely of instructional markdown and does not include any executable scripts, binaries, or external code dependencies.

prompt-optimizer