token-budget-advisor

Overview of token-budget-advisor skill

token-budget-advisor is a routing skill for choosing response depth before an answer is generated. It is best for users who want to control length, detail, or token spend instead of receiving a default-sized reply. The main job-to-be-done is simple: decide whether the next response should be short, balanced, or exhaustive, and do it early enough to avoid wasted output.

What token-budget-advisor solves

This token-budget-advisor skill helps when the user says things like “short version,” “brief answer,” “tldr,” “give me the full breakdown,” or asks to manage token usage directly. It reduces guesswork by turning vague length preferences into an explicit depth choice. That makes it more useful than a generic “be concise” instruction because it is designed to intercept and resolve the size decision first.

Best-fit users and workflows

It fits agents and workflows where answer size matters: workflow automation, interactive assistants, support triage, and any setup where long replies can be costly or noisy. token-budget-advisor for Workflow Automation is especially useful when the system needs a consistent rule for selecting answer depth based on user intent, not just a one-off prompt style.

Why install this skill

Install token-budget-advisor if your main pain point is not content quality but response sizing. It is most valuable when you need a reliable decision layer for “how much should the model say?” rather than a content-generation prompt. If the user already specified the length in the conversation, the skill should usually stay out of the way.

How to Use token-budget-advisor skill

Install and locate the source

Use the repository skill path and install token-budget-advisor from the skills/token-budget-advisor folder in affaan-m/everything-claude-code. Start with SKILL.md because that is where the trigger logic and decision flow live. If you are using the directory as a source of truth, verify the exact install path and entry file before wiring it into automation.

Turn a rough request into a usable prompt

A strong token-budget-advisor usage prompt states the task plus the desired depth choice. For example: “User wants a product summary; if they ask for options, offer short / medium / detailed before answering.” Better input also names the constraint that matters: time, cost, readability, or whether the answer will be shown inline. That gives the skill a real basis for choosing a budget.

Read first for the decision rules

For token-budget-advisor install decisions, read SKILL.md first and focus on the sections that define when to trigger, when not to trigger, and how the depth choice is made. The most useful details are the trigger phrases, the “do not trigger” cases, and the stepwise budgeting logic. Those are the parts that determine whether the skill will behave predictably in production.

Use it in a practical workflow

The best token-budget-advisor guide pattern is: detect intent, estimate needed detail, offer a depth choice if the user has not already fixed one, then answer at the selected level. That workflow works well when a system prompt or agent policy needs to preserve user control without overexplaining. Keep the prompt explicit about when to ask for a choice and when to proceed silently.

token-budget-advisor skill FAQ

Is token-budget-advisor just a prompt style helper?

No. The token-budget-advisor skill is meant to make a pre-answer decision about response depth, not merely to ask for brevity after generation has started. That difference matters in interactive systems where overshooting the budget is the actual problem.

When should I not use token-budget-advisor?

Do not use it when the user has already chosen the length in the current session, when the answer is obviously one line, or when “token” clearly means auth, payment, or session tokens. In those cases, token-budget-advisor would add friction instead of value.

Is it beginner-friendly?

Yes, if the goal is simply to manage response length. The skill is easier to adopt than a full formatting or planning system because its scope is narrow. The main learning curve is knowing when to offer depth options versus when to answer directly.

Does it fit ordinary assistant workflows?

Yes, especially where the assistant must balance brevity and completeness. token-budget-advisor is a good fit for chat interfaces, automation layers, and agents that need repeatable response-sizing rules. It is less useful if your workflow never exposes length choices to users.

How to Improve token-budget-advisor skill

Provide clearer depth cues

The best way to improve token-budget-advisor usage is to give stronger signals about what “enough” means. Instead of saying “explain this,” say “explain this in 5 bullets for a product manager” or “give a concise answer with one example.” Concrete audience, format, and length constraints make the depth choice easier and more accurate.

Watch for common failure modes

The most common failure mode is over-triggering on any mention of “token” even when the user means something else. Another is asking for a depth choice after the user already made one, which creates unnecessary back-and-forth. A third is treating every request as if it needs a budget decision, even when the answer is trivial.

Iterate after the first answer

If the first output is too short or too long, tighten the next instruction with a measurable constraint: number of bullets, approximate paragraph count, or a target audience. That feedback loop is more effective than vague edits like “more detail” or “shorter.” token-budget-advisor works best when the second turn corrects the budget, not the wording.

Make it work for your stack

For token-budget-advisor for Workflow Automation, encode the trigger conditions in your agent policy or router before generation. Keep the decision rule close to the handoff point so downstream prompts receive a clear depth target. If you are adapting the skill to another repo, preserve the trigger/do-not-trigger logic first and customize only the phrasing and budget thresholds second.