- Home
- Skills
- Codingheader
- Myskills
- 23blocks Os Ai Maestro Memory Search
23blocks-os-ai-maestro-memory-search_skill
- Python
0
GitHub Stars
2
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill codingheader/myskills --skill 23blocks-os-ai-maestro-memory-search- README.md1.2 KB
- SKILL.md4.1 KB
Overview
This skill enforces a simple but powerful default: always search conversation history before acting on any user instruction. It proactively locates previous discussions, decisions, and prior work so the assistant can continue work instead of restarting. The goal is to preserve continuity and avoid repeating or contradicting earlier conversations.
How this skill works
On every incoming user instruction the skill runs a memory search command against stored conversation content to find relevant messages, decisions, and context. It supports hybrid (default), semantic, and exact-term searches and can filter by role (user or assistant). If no relevant memory is found, the skill reports that explicitly and suggests a fallback like searching project docs.
When to use it
- Any time a user gives an instruction or requests work to be continued
- When the user references past discussions, decisions, or previous deliverables
- Before starting a new feature, bugfix, or design to avoid duplicating effort
- When the user asks for context-sensitive actions or to follow earlier constraints
- Whenever continuity and consistency across sessions matter
Best practices
- Always run a memory-search with terms from the user’s instruction before taking action
- Prefer hybrid search for general queries, semantic for concept matches, term for exact identifiers
- Check both user and assistant roles when searching for prior decisions or implementations
- If the search returns nothing, report that explicitly and then fallback to documentation searches
- Keep searches concise: use keywords, component names, or phrases the user used previously
Example use cases
- User asks “Continue working on authentication” → run memory-search.sh "authentication" before coding
- User requests to “use the approach we agreed on” → run memory-search.sh "approach" and memory-search.sh "decision"
- User reports a bug and says “like we did before” → run memory-search.sh "bug" --mode semantic and memory-search.sh "last session"
- Onboarding to a project: search for previous design notes and decisions before proposing new architecture
FAQ
Report that no prior discussions were found and propose a fallback such as searching docs or asking clarifying questions.
Which search mode should I choose?
Use hybrid for general queries, semantic when wording may differ, and term when you need exact matches like function or class names.