knowledgelisted

# Knowledge Operations ## Purpose Knowledge captures **gotchas, buried constraints, hidden invariants, and facts not in `AGENTS.md`/`CLAUDE.md`** — things a future agent would otherwise waste time rediscovering. It is not a work log, a changelog, or a duplicate of documentation that already exists. If the fact is in `AGENTS.md`, `CLAUDE.md`, or derivable from `git log`/`git blame`, do not add it. ## Contextual retrieval Do **not** dump all knowledge at session start. Instead, search for entries relevant to your current task: ```bash wade knowledge get --search "worktree safety" wade knowledge get --tag git wade knowledge get --search "error handling" --tag testing ``` When both `--search` and `--tag` are provided, entries matching **either** criterion are returned (OR semantics). Only fall back to `wade knowledge get` (no filters) if you need a broad overview or the knowledge base is small. ## Search syntax The `--search` flag accepts boolean queries: | Syntax | Meaning | |--------|---------| | `term` | Entries containing the term (case-insensitive) | | `"exact phrase"` | Entries containing the exact phrase | | `term1 AND term2` | Both terms must appear | | `term1 OR term2` | Either term may appear | | `NOT term` | Exclude entries containing the term | | `(a OR b) AND c` | Parentheses for grouping | | `worktree safety` | Bare space = implicit AND | ## Tagging ### When to tag Tag entries when they relate to specific topics that future sessions might search for. Go