context-checkup git:20260606.894c2eaA

Immutable. This exact content is served forever at /api/v1/blob/356c0d3570b7d552.

---
name: context-checkup
description: Audit what auto-loads into a Claude Code session's context window and suggest lean, reversible fixes. Use when the user asks what's loaded at startup, wants to cut context bloat or startup tokens, or keep the context window lean.
---

# Context checkup

Find what eats the session's context at startup, quantify it, and propose **reversible** trims
ranked by payoff. Read-only — measure and recommend; change settings only on explicit approval.

## Mental model (spend attention top-down)

Cost hierarchy, biggest first:
1. **MCP tool schemas** — large, but **deferred** by tool search. Only a problem if tool search is
   OFF.
2. **Always-on text** — CLAUDE.md/AGENTS.md chain, `@imports`, and **skill/agent/command
   descriptions** (every available one's description loads every session).
3. **MCP server instructions** — load per enabled server at startup; a chatty preamble can dwarf its
   tool names.
4. **MCP tool names** — ~10 tokens each. Cheap. Don't over-optimize; 30 names ≈ 300 tokens isn't
   worth a recommendation.

Rules: **measure, never guess** — `wc`/count, then attach a rough token figure to each finding.
**Rank by tokens-saved × reversibility.** Distinguish **auto-loaded** from **lazy-linked** (nested
AGENTS.md referenced from a parent are *not* loaded — don't flag them).

## Checklist

**0. Verify tool search is on first.** If `ENABLE_TOOL_SEARCH` is disabled, MCP schemas load eagerly
and every "names are cheap" conclusion flips. Check this before anything else.

**1. Always-on docs.** Measure the CLAUDE.md/AGENTS.md chain: project file + **every parent dir up
to root** + global `~/.claude/CLAUDE.md`. Expand `@import` lines (recursive). `wc -w` each.

**2. Descriptions (often the largest unexamined chunk).** Skim the in-session lists of **skills,
subagent types, and slash commands** — their descriptions are always loaded. Many rarely-used
entries = standing cost. These live in *your session's system-reminders*, not on disk — read them
from context.

**3. MCP servers.** From `.mcp.json` + settings, list enabled servers. Count tool **names** per
server (from the deferred-tools system-reminder in context). Note any server with long
**instructions**. Flag servers irrelevant to the user's actual work.

**4. Enablement levers.** Read `~/.claude/settings.json`, project `.claude/settings.json`,
`.claude/settings.local.json`. Check `enableAllProjectMcpServers`, `enabledMcpjsonServers`,
`enabledPlugins`. Precedence: managed > local > project > user. Each plugin can pull in MCP servers,
skills, agents, hooks.

**5. Memory.** Size of `MEMORY.md` index.

## Output

A short sizes table (source → ≈ tokens → auto-loaded? → relevant to user's work?), then a
prioritized action list. For each action give the exact setting/file edit, the rough savings, and
how to reverse it (prefer per-session toggles). Skip anything under ~a few hundred tokens unless
asked to be exhaustive. Then ask before editing.

## Useful probes

```bash
# always-on doc chain (run from project root; also check parent dirs + ~/.claude/CLAUDE.md)
for f in CLAUDE.md AGENTS.md ../CLAUDE.md ~/.claude/CLAUDE.md; do [ -f "$f" ] && wc -w "$f"; done
# enabled MCP + settings
cat .mcp.json; cat .claude/settings.json .claude/settings.local.json ~/.claude/settings.json 2>/dev/null
```
MCP tool-name counts, the skills/agents/commands lists, and tool-search status come from the
**session's system-reminders**, not disk — read them from your own context.