remember
$
npx mdskill add langchain-ai/deepagents/rememberCaptures conversation insights into persistent memory
- Solves preserving best practices and user feedback for future reference
- Depends on deepagents-code compatibility to store knowledge in AGENTS.md
- Decides storage location by prioritizing patterns over general observations
- Delivers results by updating memory with actionable coding conventions
SKILL.md
.github/skills/rememberView on GitHub ↗
--- name: remember description: "Review the current conversation and capture valuable knowledge — best practices, coding conventions, architecture decisions, workflows, and user feedback — into persistent memory (AGENTS.md) or reusable skills. Use when the user says: (1) remember this, (2) save what we learned, (3) update memory, (4) capture learnings." license: MIT compatibility: designed for deepagents-code --- Review our conversation and capture valuable knowledge. Focus especially on **best practices** we discussed or discovered—these are the most important things to preserve. ## Step 1: Identify Best Practices and Key Learnings Scan the conversation for: ### Best Practices (highest priority) - **Patterns that worked well** - approaches, techniques, or solutions we found effective - **Anti-patterns to avoid** - mistakes, gotchas, or approaches that caused problems - **Quality standards** - criteria we established for good code, documentation, or processes - **Decision rationale** - why we chose one approach over another ### Other Valuable Knowledge - Coding conventions and style preferences - Project architecture decisions - Workflows and processes we developed - Tools, libraries, or techniques worth remembering - Feedback I gave about your behavior or outputs ## Step 2: Decide Where to Store Each Learning For each best practice or learning, choose the right destination: ### -> Memory (AGENTS.md) for preferences and guidelines Use memory when the knowledge is: - A preference or guideline (not a multi-step process) - Something to always keep in mind - A simple rule or pattern **Global** (`~/.deepagents/agent/AGENTS.md`): Universal preferences across all projects **Project** (`.deepagents/AGENTS.md`): Project-specific conventions and decisions ### -> Skill for reusable workflows and methodologies **Create a skill when** we developed: - A multi-step process worth reusing - A methodology for a specific type of task - A workflow with best practices baked in - A procedure that should be followed consistently Skills are more powerful than memory entries because they can encode **how** to do something well, not just **what** to remember. ## Step 3: Create Skills for Significant Best Practices If we established best practices around a workflow or process, capture them in a skill. **Example:** If we discussed best practices for code review, create a `code-review` skill that encodes those practices into a reusable workflow. ### Skill Location `~/.deepagents/agent/skills/<skill-name>/SKILL.md` ### Skill Structure ``` skill-name/ ├── SKILL.md (required - main instructions with best practices) ├── scripts/ (optional - executable code) ├── references/ (optional - detailed documentation) └── assets/ (optional - templates, examples) ``` ### SKILL.md Format ```markdown --- name: skill-name description: "What this skill does AND when to use it. Include triggers like 'when the user asks to X' or 'when working with Y'. This description determines when the skill activates." --- # Skill Name ## Overview Brief explanation of what this skill accomplishes. ## Best Practices Capture the key best practices upfront: - Best practice 1: explanation - Best practice 2: explanation ## Process Step-by-step instructions (imperative form): 1. First, do X 2. Then, do Y 3. Finally, do Z ## Common Pitfalls - Pitfall to avoid and why - Another anti-pattern we discovered ``` ### Key Principles 1. **Encode best practices prominently** - Put them near the top so they guide the entire workflow 2. **Concise is key** - Only include non-obvious knowledge. Every paragraph should justify its token cost. 3. **Clear triggers** - The description determines when the skill activates. Be specific. 4. **Imperative form** - Write as commands: "Create a file" not "You should create a file" 5. **Include anti-patterns** - What NOT to do is often as valuable as what to do ## Step 4: Update Memory for Simpler Learnings For preferences, guidelines, and simple rules that don't warrant a full skill: ```markdown ## Best Practices - When doing X, always Y because Z - Avoid A because it leads to B ``` Use `edit_file` to update existing files or `write_file` to create new ones. ## Step 5: Summarize Changes List what you captured and where you stored it: - Skills created (with key best practices encoded) - Memory entries added (with location)
More from langchain-ai/deepagents
- analyze-market>-
- arxiv-searchSearches arXiv for preprints and academic papers, retrieves abstracts, and filters by topic. Use when the user asks to find research papers, search arXiv, look up preprints, find academic articles in physics, math, CS, biology, statistics, or related fields.
- blog-postWrites and structures long-form blog posts, creates tutorial outlines, and optimizes content for SEO with cover image generation. Use when the user asks to write a blog post, article, how-to guide, tutorial, technical writeup, thought leadership piece, or long-form content.
- code-reviewPerform a structured code review of changes, checking for correctness, style, tests, and potential issues.
- coding-prefsRead the user's coding preferences from /memory/coding-prefs.md before making non-trivial style decisions, and append new preferences when the user gives durable feedback.
- competitor-analysis>-
- cudf-analyticsUse for GPU-accelerated data analysis on datasets, CSVs, or tabular data using NVIDIA cuDF. Triggers when tasks involve groupby aggregations, statistical summaries, anomaly detection, or large-scale data profiling.
- cuml-machine-learningUse for GPU-accelerated machine learning on tabular data using NVIDIA cuML. Triggers when tasks involve classification, regression, clustering, dimensionality reduction, or model training on datasets.
- data-visualizationUse for creating publication-quality charts and multi-panel analysis summaries. Triggers when tasks involve visualizing data, plotting results, creating charts, or producing visual reports from analysis output.
- gpu-document-processingUse when processing large PDFs, document collections, or bulk text extraction tasks that benefit from GPU-accelerated processing. Triggers when the user provides large documents or needs bulk document analysis.