{"slug": "show-hn-token-saviour-agent-skill-from-benchmarking-9-token-saving-tools", "title": "Show HN: Token-saviour – agent skill from benchmarking 9 token-saving tools", "summary": "A developer released token-saviour, an agent skill that routes coding tasks to the most token-efficient tool based on a benchmark of nine tools, achieving up to 69.6% token reduction on comprehension tasks. The skill classifies tasks by layer—code reading, command output, prose, or code writing—and selects the best tool per layer, with validated picks including serena, rtk, caveman, and Ponytail.", "body_md": "**Spend tokens where they matter.** An agent skill that routes every coding task to the most\ntoken-efficient tool for the layer it stresses — instead of reflexively reading whole files,\ndumping raw command output into context, or writing more code than asked.\n\nToken cost has **four independent layers**, and a different tool owns each. token-saviour picks\none winner per layer — the best measured combination from a\n[9-tool benchmark](https://github.com/vagkaratzas/token-consumption-benchmark) on a realistic\nPython codebase:\n\n| Layer | Tool | Measured |\n|---|---|---|\nCode-read input (symbols, callers, call paths, architecture) |\n|\n\n**−66%** alone — the dominant cost**input**(tests, builds, git, grep, listings)[rtk](https://github.com/rtk-ai/rtk)** prose output**(chatty replies, write-ups)[caveman](https://github.com/juliusbrussee/caveman)** code output**(implementations you write)[Ponytail](https://github.com/DietrichGebert/ponytail)Stacked (serena + rtk + caveman): **−69.6% total tokens** on the comprehension suite. For\ncode-generation work, swap caveman → Ponytail (≈ −64%).\n\nThe picks were re-validated (2026-07) against newer entrants — archex, Pare, lazy-cat — and every\nincumbent held its layer. The runner-ups, the niches where they flip (e.g. Pare's structured\n`pytest`\n\nfor test-dominated loops), and the full evidence live in\n[ skills/token-saviour/references/](/vagkaratzas/token-saviour/blob/main/skills/token-saviour/references).\n\n```\n/plugin marketplace add vagkaratzas/token-saviour\n/plugin install token-saviour@token-saviour\n```\n\nOr as a plain personal skill, no plugin system involved:\n\n```\ngit clone https://github.com/vagkaratzas/token-saviour\nmkdir -p ~/.claude/skills\ncp -r token-saviour/skills/token-saviour ~/.claude/skills/\ncodex plugin marketplace add vagkaratzas/token-saviour\ncodex plugin add token-saviour@token-saviour\n```\n\nIn Codex the skill is invoked with `@token-saviour`\n\n. The VS Code Codex extension and the Codex\napp read `AGENTS.md`\n\n, which this repo ships — so it also works from the repo root with no\nsetup, and `cp AGENTS.md ~/.codex/AGENTS.md`\n\nmakes the always-on rules global.\n\nCopy `AGENTS.md`\n\n(compact, always-on ruleset) or `skills/token-saviour/SKILL.md`\n\n(full\nplaybook) into whatever instruction file your agent reads.\n\n**Classifies the task** by the layer it stresses: reading code, reading command output, writing prose, or writing code.**Routes to that layer's tool** with concrete commands (serena MCP calls,`rtk`\n\nproxies, caveman terse mode, Ponytail rules) — and degrades gracefully to plain Read/Grep/Bash when a tool isn't installed (install/verify commands in).`references/tool_links.md`\n\n**Refuses anti-patterns**: two code-read tools at once, rtk for comprehension, caveman on code, Ponytail on prose, tooling-up trivial one-line lookups.**Announces what it used**:`🪙 token-saviour: serena + rtk + caveman`\n\n.\n\n**serena**—`uv tool install -p 3.13 serena-agent`\n\n— the one code-read tool (LSP symbols; also does semantic edits/renames).**rtk**—`brew install rtk`\n\nor`cargo install --git https://github.com/rtk-ai/rtk`\n\n— add only for noisy command loops.**caveman**— see[repo](https://github.com/juliusbrussee/caveman)— add only when prose brevity is the bottleneck.** Ponytail**—`/plugin marketplace add DietrichGebert/ponytail`\n\n— add only when code-generation work is the bottleneck.\n\nAll numbers come from [token-consumption-benchmark](https://github.com/vagkaratzas/token-consumption-benchmark):\nreal tools, real artifacts, one tokenizer (`tiktoken o200k_base`\n\n), per-task tables, and honest\nlimitations (modeled-not-metered, small clean codebase, single run). Summary:\n[ references/benchmark_results.md](/vagkaratzas/token-saviour/blob/main/skills/token-saviour/references/benchmark_results.md).\n\nMIT — see [LICENSE](/vagkaratzas/token-saviour/blob/main/LICENSE).", "url": "https://wpnews.pro/news/show-hn-token-saviour-agent-skill-from-benchmarking-9-token-saving-tools", "canonical_source": "https://github.com/vagkaratzas/token-saviour", "published_at": "2026-07-08 08:49:19+00:00", "updated_at": "2026-07-08 08:59:51.285740+00:00", "lang": "en", "topics": ["ai-tools", "developer-tools", "large-language-models"], "entities": ["token-saviour", "serena", "rtk", "caveman", "Ponytail", "Codex", "Claude", "GitHub"], "alternates": {"html": "https://wpnews.pro/news/show-hn-token-saviour-agent-skill-from-benchmarking-9-token-saving-tools", "markdown": "https://wpnews.pro/news/show-hn-token-saviour-agent-skill-from-benchmarking-9-token-saving-tools.md", "text": "https://wpnews.pro/news/show-hn-token-saviour-agent-skill-from-benchmarking-9-token-saving-tools.txt", "jsonld": "https://wpnews.pro/news/show-hn-token-saviour-agent-skill-from-benchmarking-9-token-saving-tools.jsonld"}}