Memory
Skill viknesh20-20/claude-code-tool-kit/.claude/skills/memory
Production-ready Claude Code configuration. 12 original agents, 200+ slash-command skills, 45+ MCP servers, 14 plugins, design + 3D + WebGPU + GSAP + RAG tooling. One-command Node.js installer for premium websites, SaaS apps, AI agents. Free, MIT, stack-agnostic.
npx -y skills add viknesh20-20/claude-code-tool-kit --skill 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
- 5 stars5 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.
What its author says it does
Copied from the file, not written here
Manage the persistent memory layer in .claude/memory/ — list, search, save, update, delete, dedupe. Use to inspect what memories exist, write a new one, or clean up stale entries.
SKILL.md
3.1 KB, as published. Nobody here has run it
/memory
Operate on .claude/memory/. Memory persists across sessions and is shared across all 12 toolkit agents.
Subcommands
list (default)
Print the current memory index, grouped by type. One line per entry: name, description, last-verified date.
## Project (4)
- monorepo-layout — apps/, packages/, infra/. Yarn workspaces. (2026-04-18)
- billing-flag-state — webhook still off in prod, on in staging. (2026-05-01)
- migration-plan-q3 — auth migration to Lucia, target 2026-07-01. (2026-04-30)
…
## User (2)
- role — senior eng, owns the platform team. (2026-04-15)
- preferences — terse responses; no emojis; vitest over jest. (2026-04-15)
## Feedback (1)
- no-mocks-in-integration — integration tests must hit real DB. (2026-04-22)
## Reference (1)
- linear-eng-board — pipeline bugs tracked in Linear project ENG. (2026-04-15)
## Inherited (0)
## Handoffs (3)
- 2026-05-01 — auth migration shipped; billing flag pending.
- 2026-04-30 — design system v2 RFC opened.
- 2026-04-29 — initial repo onboarding.
search <query>
Find memories whose name, description, or body matches the query. Use to remind yourself of context before answering a related question.
save
Walk through writing a new memory:
- Ask which type: project / user / feedback / reference.
- Ask for the title (slug becomes the filename).
- Ask for the description (one line — used for relevance scoring later).
- Ask for the body. For
project/andfeedback/, prompt for Why: and How to apply: explicitly. - Show the file you're about to write. Get confirmation.
- Write the file.
- Update
.claude/memory/MEMORY.mdindex.
Frontmatter:
---
name: <title>
type: project | user | feedback | reference
description: <one-line>
created: 2026-05-02
last-verified: 2026-05-02
---
<body>
update <name>
Find the memory by name. Show its current contents. Ask what to change. Update the file and bump last-verified.
delete <name>
Find the memory. Show what would be deleted. Ask for confirmation. Delete the file and remove the line from MEMORY.md.
dedupe
Scan memories within each type for likely duplicates (same topic phrased differently). Surface candidates. Never auto-merge — always show pairs and ask which to keep.
Behavioral rules (enforced by the auto-loaded memory-discipline rule)
- Never write a memory the user didn't volunteer.
- Always update over add when the topic overlaps.
- Always convert relative dates to absolute.
- Never put secrets, API keys, tokens, or PII into memory.
- Always update the
MEMORY.mdindex when adding/removing.
Output format
For list, render the grouped index above. For search, show matching memories with their full body. For save / update / delete, show before/after diffs and confirm before writing.