Audit and improve prompts using FIX and REVIEW modes with validation checks.
npx skills add https://github.com/ryanthedev/oberskills --skill oberprompt
Please help me install this Agent Skill. Skill: oberprompt Source: https://github.com/ryanthedev/oberskills/tree/main/skills/oberprompt Command: npx skills add https://github.com/ryanthedev/oberskills --skill oberprompt
Prompts often behave unpredictably across models and domains; oberprompt provides a structured approach to improve reliability and safety through two modes and a rigorous workflow.
Choose a mode (FIX or REVIEW), provide the prompt, and follow the guided triage and validation steps.
High-intent search queries and answers about installing and using this skill.
To reduce flaky LLM prompt outputs, apply a structured two-mode workflow using FIX for silent improvements and REVIEW for verbose analysis, followed by a mandatory validation checklist before shipping.
The best way to audit prompts for safe behavior is to run them through a verbose REVIEW mode that enforces guardrails, checks for anti-patterns, and applies a rigorous step-by-step triage sequence.
Validate code generation prompts by applying an emergency triage sequence and a mandatory validation checklist to catch common prompting pitfalls and ensure reliable behavior across coding tasks.
Yes, you can use this dual-mode prompt engineering workflow for data analytics and research assistants, as it provides standardized checks to prevent common prompting pitfalls across these specific domains.
Prompts behave unpredictably across models due to a lack of structured guardrails and validation, which can be resolved by enforcing a dual-mode workflow that standardizes checks and prevents common anti-patterns.
The difference is that FIX mode applies silent improvements to prompts, whereas REVIEW mode provides verbose analysis, both guided by an emergency triage sequence and validation checklist to ensure reliability.
alexei-ledQuery Google Gemini via CLI for second opinions and brainstorming.
andregusman-raizDeploy code to preview, production, or rollback environments with automated smoke tests.
Zhow01Detect and block unsafe prompts and secret leakage in OpenClaw deployments.
selfagencyEvaluate AI prompts for safety, bias, security, and effectiveness.
