Prompt token optimization
Skill yigityildiz0/universal-ai-skill-library/skills/common/prompt-token-optimization
531 searchable AI Agent Skills for Claude Code, OpenAI Codex, and OpenCode — EN/TR catalog, platform and risk notes, direct ZIPs, and curated bundles.
npx -y skills add yigityildiz0/universal-ai-skill-library --skill prompt-token-optimizationAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
3 things to look at
- 18 days oldThe repository was created 18 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Reduce prompt, tool, retrieval, and conversation token overhead without weakening measured task quality. Use for context-budget audits, oversized.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
2.2 KB, 391 tokens by cl100k_base, as published. Nobody here has run it
Prompt and Token Optimization
Optimize measured waste, not length for its own sake.
Baseline
Record the task, runtime, instruction/retrieval/tool/history sizes, input/output usage when exposed by the active host/API, tool-call count, compactions, latency, cost, and quality metrics. If the host exposes no token counter, use byte/word/section estimates and say they are estimates; do not invent /cost or /usage commands.
Build a representative evaluation set including normal, boundary, adversarial, and rare high-cost failures. Preserve a baseline output and behavior checklist.
Optimization order
- Remove exact duplicates and stale platform/model instructions.
- Replace repeated prose with one decision table or invariant.
- Put trigger/selection metadata in the description and detailed material behind progressive references.
- Retrieve only relevant files/ranges; summarize evidence with pointers rather than copying entire sources.
- Reduce tool schemas and enabled tools to those needed for the current task.
- Move stable large examples/reference data to versioned files or caching when the runtime supports it.
- Compress history into facts, decisions, unresolved items, and artifact paths while preserving user constraints.
- Shorten output formatting only when it does not reduce usability or verification.
Never remove authority boundaries, security rules, acceptance criteria, uncertainty handling, or rare but severe failure checks merely because they consume tokens.
Compare
Run baseline and candidate with the same inputs and runtime settings. Measure task success, constraint violations, factual support, schema validity, tool errors, latency, tokens/cost, and variance. Revert when savings are not statistically or practically meaningful, or when any critical behavior regresses.
Deliverable
Report baseline/candidate measurements, removed/moved content, quality deltas, token/cost savings, limitations, rollout, and rollback. Do not change the selected provider/model or its reasoning level as a hidden token optimization.