patternslisted

# Project Patterns (live from the brain) The source of truth is `_dream_context/knowledge/patterns/*.md` — this skill carries NO pattern content of its own; always read the live files. 1. **No argument** → list: run `ls _dream_context/knowledge/patterns/` and show each file's `name` + first sentence of `description` from its frontmatter (Read the files, headers only). If the folder is missing or empty, say the project has no documented patterns yet and stop — do not create the folder. Ask which to load if the user's intent is unclear. 2. **With argument** → resolve `<slug or keywords>` against the filenames/descriptions (recall fallback: `dreamcontext memory recall "<keywords>" --types knowledge`), then Read the matched pattern IN FULL and apply/summarize it per the user's ask. 3. When the current work is a NEW FEATURE, always surface `feature-integration-pattern.md` — applying it before finishing is mandatory (see the dreamcontext skill's Entity Router). Do not copy pattern text into this file; do not edit patterns from here without offer-and-confirm. **You are usually not the first to know.** A pattern that the user's message actually names is already in your context: the `UserPromptSubmit` hook runs a deterministic gate and prints matching patterns with a MUST-READ directive. This skill is the BROWSE surface for the other cases — "what do we have", "load the X one". Each pattern also has its own generated `/` entry (`.claude/commands/pattern-<na