tokenarmy
/tokenarmy
Token-frugal working mode. Use for any coding or agent task to cut input and output tokens without losing accuracy.
- Author
- margesh10 ↗
- Repository
- TokenArmy ↗
- Licence
- MIT
- Last repository change
- 03.11.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 03.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Token-frugal working mode. Use for any coding or agent task to cut input and output tokens without losing accuracy.
Original author description
Token-frugal working mode. Use for any coding or agent task to cut input and output tokens without losing accuracy.
How it works
- Output (what you write)
- Input (what you load)
- Loop hygiene
SKILL.md preview
---
name: tokenarmy
description: Token-frugal working mode. Use for any coding or agent task to cut input and output tokens without losing accuracy.
---
# TokenArmy mode
## Output (what you write)
1. Lead with the answer or the action. No preamble, no recap, no closing offer.
2. No filler: skip "great question", "certainly", "I hope this helps", "let me know if".
3. Show diffs or the changed lines, not whole files. Never echo back code the user just gave you.
4. One idea per sentence, 20 words at most. Lists ≤ 5 items unless asked.
5. Keep code, paths, commands, error text and numbers exact. Brevity never applies to them.
6. Never drop meaning to save a word: keep every *not*, *never*, *no*, *only*, *except*, and every unit.
7. No text between routine tool calls. One line before a multi-step run, one line with the result.
8. If the short version is not shorter or reads two ways, write the plain sentence.
Write in full sentences when it matters more than tokens:
- security warnings and irreversible actions (confirm in full sentences first)
- steps whose order a fragment could scramble
- a confused user, or a question you ask the user
- anything that persists: code comments, commits, PRs, docs, memory files
Never rewrite or shorten the user's own prompt or quoted text.
## Input (what you load)
1. Search before reading: Grep/Glob first, then Read with offset/limit around the hit.
2. Never re-read a file you already read this session unless it changed.
3. Large file? `tokenarmy outline <file>` first, then read only the functions you need.
4. Noisy command? Use quiet flags: `git status -s`, `git log --oneline -20`, `npm i --silent`, `pytest -q`, `cargo test -q`, `--reporter=dot`. Or wrap it: `tokenarmy run -- <cmd>`.
5. Pipe long output through `| tail -n 50` or `| grep -E "error|fail"` when only the end or the errors matter.
6. Don't read lock files, minified bundles, build output, logs or `node_modules`.
7. Need a broad search across many files? Delegate to a subagent and keep only its conclusion.
8. A `[tokenarmy]` marker says what hidden lines had in common ("INFO×200; no problem lines"). Trust it; open the raw log only when you need a line it does not describe.
9. "Environment facts" at session start were learned from earlier sessions here. Use the working form they name.
## Loop hygiene
- Batch independent tool calls in one turn.
- Same failure twice? Change the hypothesis, not the retry count: re-read what the check expects, look at a different file or cause.
- Before ending a long session, write state to a file (PLAN.md) rather than relying on chat history.
Versions
The database has no other skills named tokenarmy.
Database snapshot: 13.11.2026 · package version 138672b97c6c (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100token-frugal-mode
Reduce output-token usage without losing technical accuracy. Use when the user asks to save tokens, be terse, stay brief, use caveman mode, use compressed answers, minimize chatter, or when the session is context-tight and the task does not need long prose. - 100frugal
Reduces context and token waste while preserving task-critical evidence, source pointers, and the same quality gate. Use when a software task is context-heavy, repetitive, expensive, or likely to exceed an agent context window. - 100frugal-setup
Configures the frugal plugin - which model handles each phase, which actions always need the human's approval, the session-start nudge and the token status line. Use when the user asks to set up, configure or tune frugal. - 90token-saver
Lean working mode that cuts Claude usage and plan-limit burn. Use at the start of any coding or agent session, whenever the user mentions usage, credits, limits, tokens, "save tokens", "lean mode", or "cheap mode", when a session is getting long, or when the user writes in Thai… - 90phobos
Always-on efficiency discipline — triage token spend to task size, write minimal correct code, answer terse-but-complete, take fewer round-trips, keep context lean, use memory, and route to the right tool. Use on every coding or task turn to cut token usage without losing… - 100headroom
Use when a Python or TypeScript app sends large tool outputs, logs, JSON, search results or long chat history to an LLM and the task is to cut input tokens, reduce token cost, fit a context window, or trim tool results without losing errors. Adds the headroom-ai compress() call;…