Read-only Anthropic prompt readiness workflow. It reads one Claude model, counts tokens for explicit messages and optional system prompt, then summarizes context fit, cost risk signals, and model limits. It never creates a message, starts or cancels batches, uploads/downloads files, enables tools, or calls arbitrary Anthropic API routes.

How it works 5 steps

Inspect the data fetches, transforms, gates, and output this recipe runs.

Process (5 steps)

■ model_details ?

■ token_count ?

■ message_count count

Count items in setting messages

■ summarize ?

■ card to_summary

Format results as a summary card

Settings 7 configurable

Configurable at install. Defaults shown — change them anytime in Recued.

focus setting = Review the prompt for Anthropic API readiness: input token count, model context limits, planned output budget, obvious context-window risks, and next prompt-sizing steps. Do not invent facts outside the model metadata and token count.

model setting = claude-sonnet-5

system setting =

messages setting =

anthropic setting =

max length setting = 700

planned max tokens setting = 1024