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.
npx skills add OneWave-AI/claude-skills --skill context-budgetgit clone --depth 1 https://github.com/OneWave-AI/claude-skillsWrote 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.
[](https://agentmods.dev/skills/onewave-ai/claude-skills/context-budget)<a href="https://agentmods.dev/skills/onewave-ai/claude-skills/context-budget"><img src="https://agentmods.dev/badge/skills/onewave-ai/claude-skills/context-budget/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/onewave-ai/claude-skills/context-budget"><img src="https://agentmods.dev/badge/skills/onewave-ai/claude-skills/context-budget.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00073 | $0.00886 |
| Opus 5.5 | $0.00029 | $0.00354 |
| Sonnet 5.5 | $0.00015 | $0.00177 |
| Haiku 4.5 | $0.00007 | $0.00089 |
Grade A, and why
context-budget 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 26d 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.
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.
Context Budget
Context is the scarcest resource in a long session. It gets spent on three things: files you read, output you generate, and narration nobody asked for. Only the first one is usually worth it.
Core Behavior
Treat the context window like a budget with a balance. Before any expensive action — reading a big file, running a chatty command, dumping a directory tree — ask what the cheapest way to get the same answer is.
Spend Rules
Read narrowly. sed -n '120,190p' file.ts beats reading a 2,000-line file to see one function. Grep for the symbol, then read the 40 lines around it. Read a whole file only when you are about to restructure it.
Search before reading. grep -rn "symbolName" --include="*.ts" costs a few hundred tokens and tells you which of forty files matters.
Silence the noisy commands. Pipe installs, builds, and test runs through a filter instead of dumping them whole:
npm test 2>&1 | tail -30
npm run build 2>&1 | grep -E "error|Error|warning" | head -20
npm install --silent 2>&1 | tail -5
git diff --stat # not git diff, unless you need the hunks
A passing test suite needs one line of proof, not 400.
Write instead of remembering. Anything that must survive — a plan, a decision, a list of files to touch — goes in a file on disk. Files are re-readable at a cost you choose; context is not.
Do not re-read what you just wrote. If an edit succeeded, it succeeded. Re-reading to "verify" is pure spend.
Cut the narration. No preamble, no recap of what was just shown, no bulleted summary of a diff the user can see. Say what changed and what is next, in a line or two.
Where the Budget Actually Goes
When a session bloats, it is almost always one of these:
| Leak | Fix |
|---|---|
| Whole-file reads for one function | grep, then a line-ranged read |
Full npm install / build logs |
| tail -n or grep for errors |
Directory listings of node_modules, dist, .next |
prune them in the find/ls |
| Re-reading files after editing | trust the edit result |
| Long explanations of finished work | one line |
| Pasting a file back to show a small change | show the diff hunk only |
| Repeating the plan every turn | plan lives in a file |
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.
- 26d ago First seen · 77 lines · 73 tokens per session scan A 1183ad018835
context-budget is a skill published in the GitHub repository OneWave-AI/claude-skills (348 stars, last pushed 14d ago), licensed MIT. It adds 73 tokens to every session and 886 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-09-20.
Other skills, from other repositories
magic-compact
Compact this Claude Code session.
skill-status
A study-progress companion that reads a saved JSON file to remember your learning plan, progress, predictions, and past notes between sessions.
audit-context
Measures where a project's Claude context is going before applying any of clean-context's levers. Reports what Glob discovers with and without ignore files applied, the twenty largest files in the repo and which are generated, the cumulative size of every CLAUDE.md loaded for this working directory, and how many MCP…
dream
A read-only review of project memories, notes, instructions, documents, and selected session records that produces a detailed report and a browser-viewable HTML report. It suggests ideas and insights but does not apply changes.
takenotes
Saves durable knowledge into permanent storage — typed memory, CLAUDE.md, or a docs/ file — and corrects anything already stored that has gone stale. Handles both a single fact and a whole-session harvest. Use when the user says "takenotes", asks you to remember or record something, asks you to save what was learned…
putdown
Writes a session handoff file that a fresh Claude Code session reads via /pickup, harvests durable knowledge into memory and CLAUDE.md (via /takenotes when installed), then commits and pushes all session work. Use when the user says "putdown", when ending or stepping away from a working session, or when the context…