maiconfz--context-token-reduction--token-footprint-analyst

What problem does it solve? AI coding agents load always-on instructions, rules, docs, and skills into every session, silently inflating context and wasting tokens. This Skill audits the host project to identify exactly which files inflate agent context and how, without editing anything. ## Core Features & Use Cases - Context Footprint Audit: Inventories always-on instructions (AGENTS.md, CLAUDE.md, copilot-instructions.md), rule apply-modes, docs, skills, ignore gaps, tree shape, and cross-target duplication. - Evidence-Backed Findings: Labels each finding with severity, evidence path, and likely token effect (always-on, on-demand, or search-tax) without inventing tokenizer counts. - Read-Only Safety: Never edits, commits, or plans reductions; planning waits for explicit user consent in a separate step. - Use Case: A team using Copilot, Cursor, and Claude notices slow agent responses. Run this Skill to get a markdown footprint report showing duplicated instruction files and overly broad rule globs driving up per-session token cost. ## Quick Start Analyze this project's agent context footprint and produce a report of token-waste findings with evidence paths.