ensemble-inject-standards

Ensemble Command: /ensemble:inject-standards

This Codex skill mirrors the Ensemble slash command /ensemble:inject-standards. Follow the workflow below, adapt to the current repository, and keep outputs structured.

Read the project standards index and emit a compact, task-relevant subset of coding conventions to stdout for immediate use by the current agent context. Rather than loading all standards every session, this command scores each rule against the provided task description and selects the top 5-8 most applicable rules -- keeping token cost minimal while ensuring the right conventions are surfaced at the right time.

Workflow

Phase 1: Standards Loading

1. Standards Discovery Verify the standards index exists before proceeding

• Check if standards/index.yml exists in the project root
• If standards/index.yml does not exist: print "ERROR: No standards index found. Run /ensemble:discover-standards first." and stop
• If the file exists, read standards/index.yml fully into memory
• Note the stack metadata (language, framework) and the full list of categories and rules

2. Task Analysis Parse the task description and extract keyword signals for relevance matching

• Read $ARGUMENTS as TASK_DESCRIPTION
• If TASK_DESCRIPTION is empty: print "ERROR: Provide a task description, e.g., /ensemble:inject-standards 'implement user authentication endpoint'" and stop
• Convert TASK_DESCRIPTION to lowercase and extract individual keywords
• Identify technology terms (language names, framework names, library names)
• Identify domain nouns (user, order, payment, product, auth, session, etc.)
• Identify task verbs: implement, fix, test, refactor, add, remove, update, debug, review

Phase 2: Relevance Matching

1. Category Scoring Score each standards category against the extracted task keywords