Memory is the cornerstone of intelligent agents. Without it, every interaction starts from zero. This skill covers the architecture of agent memory: short-term (context window), long-term (vector stores), and the cognitive architectures that organize them.
30
24%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./skills/agent-memory-systems/SKILL.mdQuality
Discovery
22%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This description reads like an introductory paragraph from a course syllabus rather than a functional skill description. It lacks concrete actions Claude would perform, contains no explicit trigger guidance ('Use when...'), and relies heavily on abstract conceptual language that doesn't help Claude distinguish when to select this skill.
Suggestions
Replace the abstract framing with concrete actions, e.g., 'Teaches how to implement short-term context window management, long-term vector store retrieval, and memory architectures for AI agents.'
Add an explicit 'Use when...' clause with natural trigger terms, e.g., 'Use when the user asks about agent memory, RAG, vector databases, context window management, or how to make agents remember information across sessions.'
Remove the philosophical opening sentence ('Memory is the cornerstone...') and use third-person action verbs to describe what the skill does concretely.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | The description uses abstract, conceptual language ('cornerstone of intelligent agents', 'cognitive architectures') rather than listing concrete actions Claude can perform. It reads more like a textbook introduction than a skill description. | 1 / 3 |
Completeness | The description vaguely addresses 'what' (covers architecture of agent memory) but provides no 'when' clause or explicit trigger guidance. There is no 'Use when...' or equivalent, which per the rubric should cap completeness at 2, and the 'what' itself is too vague to merit even a 2. | 1 / 3 |
Trigger Term Quality | It includes some relevant terms like 'memory', 'context window', 'vector stores', and 'agent memory' that users might mention, but these are embedded in abstract prose rather than presented as clear trigger terms. Missing common user phrases like 'remember', 'store information', 'recall', 'RAG', or 'retrieval'. | 2 / 3 |
Distinctiveness Conflict Risk | The focus on 'agent memory' and 'vector stores' provides some domain specificity, but the broad framing ('intelligent agents', 'cognitive architectures') could overlap with general AI architecture or agent design skills. | 2 / 3 |
Total | 6 / 12 Passed |
Implementation
27%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill reads more like a comprehensive tutorial or reference guide than a concise, actionable skill file. While it contains genuinely useful patterns (contextual chunking, memory decay, conflict resolution) and good sharp edges documentation, it is far too verbose for a SKILL.md, explains many concepts Claude already understands, and fails to organize content across multiple files. The code examples are mostly concrete but inconsistently formatted and sometimes rely on undefined helper functions.
Suggestions
Reduce the SKILL.md to a concise overview (~100 lines) with core principles and the most critical pattern (memory type architecture), then split vector store examples, chunking strategies, sharp edges, and validation checks into separate referenced files.
Remove explanatory content Claude already knows (what memory types are, what vector databases do, 'The chunking dilemma' explanation) and replace with terse decision rules.
Move the Capabilities, Scope, When to Use, Limitations, and Collaboration sections to YAML frontmatter or remove them - they are metadata, not actionable instructions.
Fix code formatting: replace triple-quoted strings with proper markdown code blocks, and ensure examples use defined imports/functions or clearly mark dependencies.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | This is extremely verbose at ~600+ lines. It explains concepts Claude already knows (what memory types are, what vector databases do), includes extensive tooling comparison tables, and repeats patterns like contextual chunking multiple times. The 'Capabilities', 'Scope', 'When to Use', and 'Limitations' sections are metadata that belong in frontmatter, not body content. Much of this reads like a tutorial rather than actionable instructions. | 1 / 3 |
Actionability | The code examples are mostly concrete and near-executable (LangMem, Pinecone, Qdrant, ChromaDB, chunking strategies), but many are wrapped in triple-quoted strings rather than proper code blocks, some use undefined functions (summarize, embed, store, cluster_by_similarity), and several patterns are pseudocode-level. The LangMem API examples may not reflect actual library interfaces accurately. | 2 / 3 |
Workflow Clarity | Individual patterns are reasonably well-structured with 'When to use' annotations and the Sharp Edges section provides good error-recovery guidance. However, there's no clear end-to-end workflow for implementing a memory system - it's a collection of patterns without a sequenced process. Validation checks are listed but not integrated into workflows as checkpoints. | 2 / 3 |
Progressive Disclosure | This is a monolithic wall of text with no references to external files despite being 600+ lines. All content - tooling comparisons, multiple vector store implementations, chunking strategies, sharp edges, validation checks - is inlined. There are no bundle files, yet the content desperately needs to be split into separate reference documents (e.g., vector store comparison, chunking guide, sharp edges). | 1 / 3 |
Total | 6 / 12 Passed |
Validation
81%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 9 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
skill_md_line_count | SKILL.md is long (1089 lines); consider splitting into references/ and linking | Warning |
frontmatter_unknown_keys | Unknown frontmatter key(s) found; consider removing or moving to metadata | Warning |
Total | 9 / 11 Passed | |
8854d4e
Table of Contents
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.