docmancer-memory

# docmancer memory Executable: `docmancer` Docmancer turns memory and instructions scattered across coding agents into one curated, source-attributed Markdown tree. Harvested files remain read-only evidence, ambiguous capture stays in the inbox, and curated Markdown files are the canonical memory every supported agent can read. ## Workflow 1. Run `docmancer init` to create or adopt the current project's tree. This never enables capture implicitly. 2. Run `docmancer context "task" --project-path "$PWD"` before work when prior decisions or conventions may matter. 3. Use `docmancer read <address>` to inspect the complete Markdown file and provenance before changing it. 4. When the user explicitly asks to remember a durable fact or decision, use `docmancer write` with an explicit project-relative destination. 5. Use `docmancer harvest <path>` to preview harvested evidence. Add `--apply` only when the user wants it copied into the uncurated inbox. 6. Use `docmancer curate` to preview a complete file diff. Add `--apply` only after the destination and content are acceptable. 7. Keep Docs separate with `docmancer docs ...`. ## Commands ```bash docmancer init --project-id <stable-project-id> docmancer context "what deployment decisions apply?" --project-path "$PWD" docmancer search "deployment" docmancer read docmancer://memory/<id> docmancer write "# Release process\n\nDeploy on Railway." --path deployment/release.md --scope project --project-id <id> docmancer edit docmancer://...