{"slug": "i-created-nitro-skills-to-reduce-token-usage-of-my-claude-agents", "title": "I created Nitro Skills to reduce token usage of my Claude agents", "summary": "Dextrata released Nitro Skills, an open-source toolkit of 15 skills that reduces token usage for Claude agents by keeping each skill's logic in an executed script rather than in the model's context. Each skill is a short SKILL.md file plus a script, and the package requires only Python 3, installing into the global Claude directory and wiring up supporting hooks automatically via a git clone of the Dextrata/nitro-skills repository followed by python install.py. The skills include hashpatch for editing files through hashed line anchors, rerun for caching test, build, and lint output, and budget for tracking token spend per command and blocking unbounded floods.", "body_md": "Each skill is a short `SKILL.md` — the only part that ever enters the model's context — plus a script that gets executed, never read. The clever logic costs zero tokens no matter how many times it's reused.\n\nThe first three shrink reads, edits and repeated command output. The rest attack everything else: repeated output, long strings, reassurance reads, mechanical edits, logs, payloads, traces, diffs, search hops, re-derived facts, and unbounded floods.\n\n`hashpatch`\nEdit files by pointing at hashed line anchors instead of quoting whole files.\n\n`rerun`\nCache and reuse output from tests, builds, and lints across repeat runs.\n\n`probe`\nInspect a module's API by running it, not by reading its source.\n\n`seen`\nFold already-shown output blocks into one-line pointers.\n\n`alias`\nShorten long repeated strings (hashes, UUIDs, paths) into small tokens.\n\n`believe`\nVerify a claim about a file against reality without re-reading it.\n\n`refactor`\nDescribe a mechanical multi-file edit in a sentence instead of a diff.\n\n`mine`\nCluster huge logs into a handful of templated lines with counts.\n\n`shape`\nPrint the schema of structured data instead of dumping the payload.\n\n`trace`\nStrip library noise from stack traces down to the user-code frames.\n\n`sdiff`\nSummarize a diff by symbol instead of printing every changed line.\n\n`q`\nAnswer structural code questions from a cached symbol index.\n\n`blast`\nShow a symbol's callers and callees before you change its signature.\n\n`recall`\nPersist answers to \"where/how is X handled\" across sessions.\n\n`budget`\nTrack token spend per command and block unbounded floods.\n\nRequires only Python 3. Installs into your global Claude directory and wires up the supporting hooks automatically.\n\n```\ngit clone https://github.com/Dextrata/nitro-skills.git\ncd nitro-skills\npython install.py\n```\n\nSee the [full README](https://github.com/Dextrata/nitro-skills#readme) for manual install, flags, and a `CLAUDE.md` example that makes usage legible to the agent.", "url": "https://wpnews.pro/news/i-created-nitro-skills-to-reduce-token-usage-of-my-claude-agents", "canonical_source": "https://nitroskills.com", "published_at": "2026-09-12 23:34:51+00:00", "updated_at": "2026-09-12 23:55:36.913322+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "ai-products"], "entities": ["Nitro Skills", "Dextrata", "Claude", "Python 3", "hashpatch", "rerun", "budget"], "alternates": {"html": "https://wpnews.pro/news/i-created-nitro-skills-to-reduce-token-usage-of-my-claude-agents", "markdown": "https://wpnews.pro/news/i-created-nitro-skills-to-reduce-token-usage-of-my-claude-agents.md", "text": "https://wpnews.pro/news/i-created-nitro-skills-to-reduce-token-usage-of-my-claude-agents.txt", "jsonld": "https://wpnews.pro/news/i-created-nitro-skills-to-reduce-token-usage-of-my-claude-agents.jsonld"}}