atlas-compact

atlas-compact is a skill for Claude Code from Blazity/atlas. It costs 49 tokens per session (984 once invoked), scanned A, original, MIT.

A skill for reducing oversized instructions, memory files, and other documents that an AI coding agent reads.

In plain words
What is it for?
Use it when Atlas reports context-size problems or when files such as AGENTS.md or CLAUDE.md have become too large.
Why use it?
It helps remove duplication, move less important material, and keep the agent’s working context manageable without guessing at measurements or thresholds.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: mentions CLAUDE.md; mentions AGENTS.md.

Part of the atlas plugin — 3 skills shipped together

Good fit Use it when Atlas reports context-size problems or when files such as…

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/blazity/atlas/atlas-compact
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add Blazity/atlas --skill atlas-compact
Clone the repo
git clone --depth 1 https://github.com/Blazity/atlas

Made for: Claude Code.

Or install atlas, the plugin that ships this one along with the rest of its 3 skills.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for atlas-compact

README.md
[![agentmods](https://agentmods.dev/badge/skills/blazity/atlas/atlas-compact.svg)](https://agentmods.dev/skills/blazity/atlas/atlas-compact)
Your own site
<a href="https://agentmods.dev/skills/blazity/atlas/atlas-compact"><img src="https://agentmods.dev/badge/skills/blazity/atlas/atlas-compact.svg" alt="Measured on agentmods" height="20"></a>
Per session 49 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 984 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00049 $0.00984
Opus 5 $0.00024 $0.00492
Sonnet 5 $0.00010 $0.00197
Haiku 4.5 $0.00005 $0.00098

Measured 7d ago against content hash 468185810fb0, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

atlas-compact scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.ai/skills/atlas-compact/SKILL.md · 77 lines

How it starts

The opening of the file, as written. The whole thing — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Compact AI Context

Overview

Use this skill when atlas doctor reports context-size advisories or AI-facing context files grow oversized. The published CLI owns deterministic measurement — which files count, their sizes, thresholds, and statuses. This skill owns semantic judgment — what to cut, move, merge, or keep. Never reimplement measurement, guess sizes, or invent thresholds.

Call the CLI through npx, preferring a locally installed @blazity-atlas/core (npx --no-install @blazity-atlas/core …) and falling back to the published package only when none is installed. Every npx --yes @blazity-atlas/core@latest … command in this skill is the fallback spelling of that rule, not an instruction to skip the local copy.

This skill edits content only. Structure repair — directories, config, symlinks, managed files — stays with doctor --fix.

Run the phases below in order.

Phase 1 — Deterministic Measurement

From the repository root, run:

npx --no-install @blazity-atlas/core doctor

Follow the result before compacting anything:

  • Manual findings (exit 2): summarize the conflicts and stop.
  • Fixable drift (exit 1): offer npx --yes @blazity-atlas/core@latest doctor --fix first — structure repair precedes compaction. If --fix refuses because of a dirty worktree, ask the user whether to commit, stash, or explicitly rerun with --force; never use --force automatically.
  • No context-size advisory and no user-named target: report that the AI context is within thresholds and stop.

Capture the authoritative per-file worklist:

npx --no-install @blazity-atlas/core doctor --handoff context-size

Treat its entry lines as the worklist; do not re-measure by hand.

Phase 2 — Read and Classify

Resolve every path through the workspace config — never hardcode .ai/: read .ai/config.json, or when absent, follow the .atlas repo-root pointer to <root>/config.json.

Read each flagged file and bucket its content:

  • Commands, invariants, and safety rules → stay in root instructions.
  • Durable product, architecture, stack, and lesson detail → configured memory files.
  • Decisions and rationale → configured decisions/ADR paths.
  • Canonical terms → the configured language file.
  • Duplicated content → keep one canonical location and reference it from the rest.
  • Stale content contradicted by the repository (with evidence) → delete.

Read the full file on GitHub · 77 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 7d ago First seen · 77 lines · 49 tokens per session scan A 468185810fb0

Subscribe to this mod's changes

atlas-compact is a skill published in the GitHub repository Blazity/atlas (24 stars, last pushed 1mo ago), licensed MIT. It adds 49 tokens to every session and 984 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.