prompt-master

State the task, target tool, and desired output format explicitly, then let the Skill route to the correct model guidance. It produces a single copyable prompt block using compact sections like Goal, Context, Constraints, and Done.

Paste the existing prompt and the Skill scans it against 37 documented failure patterns such as vague task verbs, missing success criteria, and absent stop conditions. It returns a rewritten version with the gaps filled.

Yes. Midjourney needs comma-separated descriptors with parameters like --ar and --v at the end, while DALL-E 3 accepts prose descriptions. The Skill applies the correct syntax automatically based on the target tool.

Yes, the Prompt Decompiler mode handles adaptation. Specify the source and target tools, and it rewrites the prompt using the target tool's syntax and best practices while listing the key changes made.

They must include starting state, target state, file scope, forbidden actions, stop conditions, and human review triggers. The Skill enforces these through Templates G, H, and M to prevent runaway agents and scope creep.

Avoid them for reasoning models like o3, DeepSeek-R1, and Qwen3 thinking mode, where added scaffolding degrades output. The Skill instead requests conclusions, assumptions, evidence, and verification checks.