skills.rest
CleanExpo
CleanExpoCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

tao-context-mode

Build compact per-file repository summaries with regex-based indexing and on-demand expansion.

Updated Aug 23, 2026
Search Tags:#context compression#codebase indexing#on-demand retrieval#TAO orchestration#deterministic summarization#GitHub repository analysis
One-click install
npx skills add https://github.com/CleanExpo/Pi-Dev-Ops --skill tao-context-mode
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: tao-context-mode
Source: https://github.com/CleanExpo/Pi-Dev-Ops/tree/main/skills/tao-context-mode
Command: npx skills add https://github.com/CleanExpo/Pi-Dev-Ops --skill tao-context-mode

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It solves the problem of overwhelming LLM context windows by forcing agents to choose between shallow repo awareness and expensive full-file reads.

Core Features & Use Cases

  • Summary-indexing of repositories: Builds a compact CodebaseIndex with per-file summaries and lightweight symbols by walking the repo once.
  • On-demand expansion via expand(path): Returns the full file contents only for paths explicitly needed for the current task.
  • Deterministic, low-cost synopsis: Uses pure regex and first-comment synopsis to avoid LLM calls while supporting sha256-keyed invalidation for repeat runs.
  • Operational use in TAO loops: Fits planner and agent iteration workflows by providing fast “where does X live?” lookups and reducing token pressure during successive iterations.

Quick Start

Ask your agent to build the index for your repo root and then expand the specific file path it needs for the current task (for example, expand("app/server/main.py")).

Frequently Asked Questions about tao-context-mode

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reduce token usage when analyzing a GitHub repository?▼

On-demand codebase retrieval serves compact per-file summaries instead of full source contents, using an expand(path) mechanism to return only the requested file content and drastically reduce token pressure during repository analysis.

How does deterministic regex-based codebase indexing work?▼

Deterministic regex-based codebase indexing works by walking the repository once to extract lightweight symbols and first-comment synopses without relying on LLM calls, using sha256-keyed invalidation to maintain cache validity on repeat runs.

Can I use context compression for TAO loop session planning?▼

Yes, context compression fits TAO loop session planning by providing fast "where does X live?" lookups and reducing token pressure, allowing planner and agent iteration workflows to maintain codebase awareness efficiently.

What is the best way to find where specific code lives in a large codebase without exceeding context limits?▼

The best way to find where specific code lives without exceeding context limits is querying a compact CodebaseIndex of per-file summaries, which provides fast codebase awareness and uses on-demand expansion to fetch only the needed full file contents.

Does this repository analysis approach require LLM calls for generating file summaries?▼

No, this repository analysis approach uses deterministic, low-cost synopsis generation based on pure regex and first-comment extraction, avoiding LLM calls entirely while supporting sha256-keyed invalidation for repeat runs.

Related Skills

Explore More Software Engineering Skills→
bibo1243bibo1243

LINE Local Reader

Extract concise summaries from local LINE HTML chat backups.

Community
Intermediate
NVIDIA-TAONVIDIA-TAO
83

tao-run-deft-aoi

Automate the DEFT AOI improvement loop for PCB inspection models.

Official
Advanced
maestria-comaestria-co

code-identifier

Locate code implementations and entry points across a codebase.

Official
Intermediate
laurentvvlaurentvv

Utilisation directe des Tools Python pour accélération

Execute Oil-Agent Python tools directly to bypass the smolagents cycle.

Community
Intermediate