Token Efficiency

Avoid unnecessary context burn. Surgical file reads, batched commands, less narration. Every token read is a token taxed.

Safety Notice

This listing is from the official public ClawHub registry. Review SKILL.md and referenced scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "Token Efficiency" with this command: npx skills add nubzparmesan/minnow-token-efficiency

Token Efficiency - Avoiding Unnecessary Context Burn

Every token read is a token taxed. Be surgical.

Core Rules

  • grep/tail over full reads — need one section? grep -A 20 "## Section". Need the end? tail -50. Don't read 500 lines to find 5.
  • batch API calls — one curl with multiple params > three separate curls. Combine endpoint checks into single scripts.
  • sub-agents for heavy research — scraping a thread, summarizing 50 posts, reading a full repo? Spawn a sub-agent. Don't burn main context on bulk work.
  • don't re-read context — if a file was already read this session, use what you have. Never read it again "just to be sure."
  • summarize before storing — tool output going to a file? Strip it down first. Store the insight, not the raw dump.

Anti-Patterns (Observed)

BadGood
Read full 3000-line comment threadgrep -c "" for count, tail for recent, grep for keywords
4 separate curl calls to 4 endpointsOne script, four curls, one python parse pass
Narrate every grep and read stepJust do it, report the finding
Re-read SKILL.md to "refresh" mid-sessionTrust what's already in context
Store raw JSON API responseExtract the 3 fields you care about, store those

When to Use Sub-Agents

  • Research tasks > ~10 tool calls
  • Bulk reading (threads, repos, feeds)
  • Tasks that can run in parallel with main work
  • Anything that would dominate your context window

Quick Heuristics

  • If you're about to read a file you've already read → stop
  • If you're about to make the same API call twice → cache the result
  • If narrating a step takes longer than doing it → just do it
  • If a task feels heavy → sub-agent it

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

General

地藏经药师经智慧

地藏经药师经智慧 - 佛家孝道与救度思想,涵盖地藏本愿、药师十二愿、因果报应、消灾延寿等核心智慧,适用于道德修养、慈悲精神、身心健康

Registry SourceRecently Updated
General

Precision Oncology Zhcn

综合学术文献、流行病学报告、临床与药物指南及临床试验报告,提供关于癌症及其治疗的报告。 基于癌变机制进行详细的分子生物学和组织学分析。 当查询涉及以下内容时加载本技能: - 癌症或肿瘤 - 癌变机制 - 癌症或肿瘤的治疗 典型查询 - 乳腺癌是如何发生的? - 白血病的一线和二线治疗 - CAR-T 疗法治疗胰腺...

Registry SourceRecently Updated
General

hermes-traffic-guardian

Hermes runtime traffic monitoring baseline for opt-in proxy inspection, egress detection, and attestation-aware traffic posture.

Registry SourceRecently Updated
General

Scp Paradigm

Use when analyzing how industry structure drives firm behavior and market performance, assessing market concentration, entry barriers, or competitive dynamic...

Registry SourceRecently Updated