Auto memory
Auto-save context when the agent learns something — triggers on: remember, convention, always do, note for future, learned that, from now on, never do, experiment completion, bug resolutionFrom its SKILL.md
npx -y skills add Gaaaavin/claude-eureka --skill auto-memoryAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
1.7 KB, as published. Nobody here has run it
When to save
Activate when any of these occur:
- User says to remember something, establishes a convention, or sets a preference
- An experiment completes and results are logged
- A bug is resolved and root cause is identified
- A significant architectural discovery is made about the codebase
Where to save
Route entries to the appropriate context file:
| Content type | File |
|---|---|
| Conventions, preferences, workflows | .claude/context/conventions.md |
| Experiment results, metrics, findings | .claude/context/experiments.md |
| Architecture discoveries, code patterns | .claude/context/architecture.md |
Create the file if it does not exist. Use a # Context heading at the top.
Entry format
Each entry is a dated, concise line:
- **YYYY-MM-DD**: <short description of what was learned or decided>
Append new entries under the appropriate section heading. Keep descriptions to one or two sentences.
Deduplication
Before adding an entry:
- Read the target context file
- Search for keywords from the new entry
- If a substantially similar entry exists, update it in place instead of adding a duplicate
- If updating, preserve the original date and append the new date:
(updated YYYY-MM-DD)
Guidelines
- Be concise — context files are loaded into every conversation
- Only save things with lasting value, not one-off facts
- Never store secrets, tokens, or credentials
- Prefer actionable entries ("use X instead of Y") over observations