Create a memory from natural language or structured flags
View on GitHubplugins/ccmem/skills/ccmem-remember/SKILL.md
February 3, 2026
Select agents to install to:
npx add-skill https://github.com/gaodes/cc-ecosystem/blob/main/plugins/ccmem/skills/ccmem-remember/SKILL.md -a claude-code --skill ccmem-rememberInstallation paths:
.claude/skills/ccmem-remember/# /ccmem-remember Create a new memory to remember user preferences, patterns, or workflows. ## Usage ``` /ccmem-remember <natural language description> /ccmem-remember --type <type> --title "<title>" [options] ``` ## Arguments The command accepts either natural language or structured flags: **Natural language:** Parse the input to extract type, title, and description. ``` /ccmem-remember prefer pnpm over npm /ccmem-remember always run tests before committing ``` **Structured flags:** - `--type <type>` — preference, pattern, workflow, correction (default: preference) - `--title "<title>"` — memory title - `--description "<desc>"` — detailed description - `--confidence <0.0-1.0>` — initial confidence (default: 0.7) - `--project` — scope to current project only - `--tag <tag>` — add a tag (can repeat) ## Execution Steps 1. Parse the input (natural language or flags) 2. For natural language, infer: - Type: "prefer" → preference, "always/never" → pattern, "workflow" → workflow - Title: Extract the core preference/pattern - Description: The full input 3. Run `ccmem create` to create the memory interactively, providing the parsed values 4. Display the created memory with its ID and confidence 5. Suggest: "Use `/ccmem-reinforce <id>` if this works well for you" ## Examples **Natural language:** ``` User: /ccmem-remember prefer rg over grep for searching → Creates preference memory: "Prefer rg over grep for searching" ID: 2026-02-03T14-30-00-prefer-rg-over-grep Confidence: 0.7 ``` **With flags:** ``` User: /ccmem-remember --type workflow --title "Test before commit" --project → Creates project-scoped workflow memory ``` ## CLI Command ```bash ccmem create ``` Runs interactively; provide parsed values when prompted.