Audits module architecture and ranks refactoring candidates by evidence read from scoped code.
npx skills add https://github.com/blauwtje/exo --skill deepen-blauwtje
Please help me install this Agent Skill. Skill: deepen Source: https://github.com/blauwtje/exo/tree/main/skills/deepen Command: npx skills add https://github.com/blauwtje/exo --skill deepen-blauwtje
scripts/hotspots.mjs. - Delegated evidence-based analysis: Dispatches a general-purpose delegate on Opus that reads the scoped code within a strict read budget and writes at most 5 ranked cards with friction, refactor, payoff, confidence, and migration cost. - Two output modes: Produces ranked cards in the reply, or in a read-only planning mode writes a full plan artifact following the planning skill's plan specification. - Use Case: Ask where the coupling problems are in your checkout flow, and receive ranked cards naming exact files and symbols, each with the interface the refactor leaves and its migration cost. ## Quick Start Ask the assistant to run the deepen skill on the payments module to find the refactors that would give it a smaller interface over its implementation.High-intent search queries and answers about installing and using this skill.
Run an architecture audit scoped to a path or module, or let the hotspots script rank the paths committed to most in the last six months. A delegate reads the scoped code and writes at most 5 ranked cards, each naming files, friction, the refactor, payoff, confidence, and migration cost.
A module is shallow when learning its interface teaches most of its implementation. The audit applies the deletion test: imagine inlining the module, and if complexity gathers in one place it is shallow; if it only spreads, the claim is recorded as friction instead.
Yes. In a read-only planning mode or when a plan is requested, the skill audits and then writes a plan for the top findings into the harness's plan file or docs/plans/<topic>.md, following the planning skill's plan specification. Otherwise it returns ranked cards with no edits.
Do not use it for a single named refactor or rename, which is a decided change for a refactoring skill, nor for failing existing behavior, which routes to debugging. It also does not perform security or dependency audits.
The delegate writes the ranked cards to <git dir>/exo/deepen/<topic>.md, which acts as a compaction anchor. After a compaction, the session rereads that file rather than rerunning the audit or re-reading the scoped code.
josippapezGuides one-question-at-a-time dialogue to define outcomes and approaches before coding.
limaonEvaluate code changes for quality and compliance before merging.
aicodeproDiagnose software bugs through evidence-first root-cause analysis and minimal fixes.
tonehqOrchestrate renames, extractions, and moves across codebases with GitNexus workflows.