Token saving

Usage limits depend on how much text goes through the model. agent-skills keeps that small in five ways. The numbers below were measured on real projects by the agent-skills authors.

An agent keeps only each skill’s name and description in context. The full SKILL.md loads when the skill runs, and reference files only when a step needs them. Every SKILL.md is at most 110 lines, and a check in CI enforces it.

Fetching, parsing, counting and filtering are done by scripts that print only what the model needs. Sizes are in bytes, and four bytes are roughly one token.

Task	Raw input	Script output	Saved
Lighthouse report	455,534	754	99.8%
Test log of 86 passing tests	27,098	86	99.7%
npm audit --json on a 300-package app	20,555	2,126	90%
git log -60 to learn the commit style	7,415	739	90%
GitHub issue and comments	8,178	1,018	88%
SQL log of one request with an N+1	4,204	490	88%
package.json to find test and lint commands	2,122	275	87%

Fetching a task again prints only new comments and changed fields. Failing test runs print the failing blocks with file:line and a summary, never the full log.

A screenshot costs about 1,000 to 2,000 tokens, so browser-check asks first and captures only the changed element. Local Lighthouse, full test suites and load tests are proposed, not run silently.

STATE.md, about 30 lines long, holds the goal, what was done, changed files, check results, decisions and one concrete next step. After a stage, clear the session and resume from the file. It is cheaper and more precise than dragging a long transcript along.

The token-saver skill gives every agent the same habits:

• Search before reading, and read line ranges.
• Never open lockfiles or build output.
• Use short command forms, such as git status --short.
• Edit instead of rewriting.
• Put results first in answers.
• Use subagents with narrow briefs for wide searches.

• Skills catalog: all skills.
• Install agent-skills: set up the skills.
• Features: all views of Skribly Agent, grouped by job.