# Context Mode

> Context optimization and compression routing rules for AI agents exploring large codebases, reading massive logs, searching symbols, and batching tool executions.

- Skill: `pedroiff0/context-mode` (Agent Skill)
- Install (CLI): `npx skillmds@latest add pedroiff0/context-mode`
- Raw SKILL.md: https://api.skillmd.com/api/skills/pedroiff0/context-mode/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: MIT
- Author: pedroiff0 (https://skillmd.com/u/pedroiff0)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/pedroiff0/context-mode

---


# Context Mode & Token Optimization

This skill provides operational patterns for navigating massive repositories while conserving context tokens using indexed searches, AST filters, and structured outputs.

## When to Use

- Reading large log files (>500 lines) or analyzing multi-megabyte code repositories.
- Performing symbol lookups across hundreds of files without flooding context.
- Running high-volume batch executions or lint sweeps.

## Operational Guidelines

1. **Targeted Line Range Views**: Avoid dumping entire files; view slices using line offsets (`view_file` with `StartLine` and `EndLine`).
2. **Grep Before View**: Use ripgrep (`grep_search`) to locate exact function definitions or usages before loading the containing file.
3. **Subagent Offloading**: For deep exploratory research, invoke an isolated research subagent (`invoke_subagent`) to read docs and summarize findings back to parent context.