claude

scratch-memory

Project-scoped scratch-memory: MCP tools (write_report, write_review, write_issue, write_session, write_task) for sub-agent verdicts, session handoffs, and task capture plus scratch-memory CLI (handoff, pickup, register with two opt-in PostToolUse hooks, tasks, epics) for session state persistence and ownership transfer across Claude Code sessions, plus the commit-session verb for per-session file validation and rewrite-pointer/cat-sessions (with --with-tasks) verbs for v3 HANDOFF.md pointer regeneration and resume briefs; plus the capture-issue command and the scratch/issues/ corpus conventions (schema including the optional epic/spike keys, open|resolved lifecycle, triage, resolution) documented in scratch-issues-methodology, with the epics verb group deriving an epic's workable spike frontier and its whole member graph from those files; plus the wrap-up skill and its wrap collect verb, which sort a finished workstream's decisions, lessons, open questions, tasks and stray files into durable destinations before archiving it

0 on GitHubby Richard Adletav1.3.4MITupdated 22 days agoproductivity
/

Plugin scratch-memory from marketplace radleta in radleta/claude-plugins. Commands read from the repo's manifest — 19 days ago.

What it installs

Skills (12)

  • commit-methodology

    Comprehensive commit creation with adaptive analysis, security screening, conventional format, and execution protocol. Use when creating git commits, analyzing staged changes, screening for secrets, or formatting commit messages — even for quick single-file commits.

  • design-record

    Design-record rubric: the decision line (X, not Y, because Z) and the form each kind of design content takes, with diagram rules, anti-bloat tests and healing rows. Use when writing or reviewing a decision, an epic, an idea.md, an issue or a wiki page's design — even one decision line.

  • handoff-methodology

    Handoff and resume: per-session files, workstream tasks, the HANDOFF.md pointer, the /pickup brief, task grooming. Use when writing a handoff, picking up a session, creating or editing a workstream task, grooming a workstream, handling a legacy HANDOFF.md folder, or choosing skills to re-load — even

  • knowledge-capture

    Knowledge capture layer for `learned/` files — capture heuristic, frontmatter schema, slug derivation, drift severity, capture checkpoint, boundary ingest. Use when writing a learned file, checking learned/ status, running a boundary ingest, or evaluating a discovery at a pause — even when nothing s

  • knowledge-distillation

    Systematic admission gate for evaluating which domain knowledge earns inclusion in expert skills — the four admission questions plus the principle-versus-instance filter, knowledge categories, and compaction techniques. Use when distilling insights from research, curating knowledge for skill creatio

  • llm-wiki-expert

    LLM-maintained personal wiki pattern — persistent, compounding knowledge bases built from interlinked markdown instead of RAG. Use when setting up a personal wiki, designing wiki schemas, ingesting sources, or maintaining wiki health — even for simple note organization or when mentioning Obsidian or

  • scratch-issues-methodology

    Conventions for the scratch/issues/ corpus — frontmatter, statuses, severity, queries, filing, resolution, and the issue standard (admission, groom determinations, fold-first build sizing). Use when reading, querying, filing, grooming, sizing builds from, rating or closing anything in scratch/issues

  • scratch-management

    Scratch folder lifecycle management using scratch.mjs script and archive branch workflow. Use when saving, archiving, initializing, or cleaning up scratch folders, managing agent context visibility, or hiding completed planning docs — even for simple single-folder operations.

  • scratch-memory

    Write-only scratch/ persistence: MCP write tools for restricted sub-agents and a CLI. Use when registering the MCP, hooks or repos, running /handoff or /pickup, wiring a verdict, capturing, closing or linting issues and tasks, deriving an epic's spike frontier, or collecting a workstream for wrap-up

  • user-decision-rubric

    Rubric for whether a call at a workflow stop is the user's or the agent's, and how a ruling is recorded. Use when writing or editing any skill, command or agent that stops for the user, lists rulings for approval, or escalates — even when the stop site has its own owner line.

  • wiki-memory

    Methodology and scripts for wiki-backed expert skills — the page standard every writer applies, the ingest-groom rubric, the whole-domain sweep, and the init, lint, query, audit and status protocols. Use when filing captured knowledge onto a wiki page, grooming or auditing a domain, querying an expe

  • wrap-up

    Wrap up a workstream before archiving: sort each decision, lesson, question, task or stray file in scratch/S-{id}/ into a durable home, land what the user approves, commit, archive. Use when a workstream is finished, when asked to close or archive an S- folder, or when resuming a stopped wrap-up — e

Slash commands (10)

  • /scratch-memory:capture-issue

    Capture an issue, feature idea, or mixed pain-point as a structured markdown file in scratch/issues/

  • /scratch-memory:commit-all

    Multi-repo commit: surveys all sub-repos via .subrepos and creates coordinated conventional commits. Use when committing work that spans the main repo, scratch or another sub-repo — even when only one repo has changes.

  • /scratch-memory:commit

    Create conventional commit with security checks, smart staging triage, and proper formatting

  • /scratch-memory:groom

    Groom a workstream's tasks to the current task methodology — dispatch the employee agent against scratch/S-{id}/, then apply its user-only edits, for veto on a lane and on your confirmation off one. Use when /pickup prints a GROOM: line or an aged workstream's tasks need bringing up to date — even f

  • /scratch-memory:handoff

    Save the current session via write_session and reconcile owed work into workstream tasks; the HANDOFF.md pointer regenerates from the session log and the close ledger. Use when pausing or ending a session whose work continues, or before /clear — even when little was done.

  • /scratch-memory:ingest

    Run the knowledge-ingestor over a project's learned/ captures, or groom a wiki domain

  • /scratch-memory:pickup

    Resume a prior session by assembling a resume brief via cat-sessions, running any instruction given, then executing Next best step

  • /scratch-memory:start

    Start a new workstream by writing its first handoff session under a new slug

  • /scratch-memory:wiki-init

    Create a new wiki-backed skill folder for a domain, scaffolding its SKILL.md, schema.md and .mditerc so the ingestor and researcher can file into it

  • /scratch-memory:wiki-status

    Show pending and escalated learned/ captures per project and the wiki fleet's health, or one domain's pages and state

Subagents (6)

  • commit-worker

    Creates well-crafted conventional commits with smart staging triage. Use when committing code, creating git commits, or formatting commit messages with security screening.

  • employee

    Standard-tier generic worker that does exactly what its dispatch prompt states. The default for delegated work (building, editing, sorting, researching, and high-stakes calls such as security or release gates), and for re-checking a verdict `intern` got wrong, even a one-line task. A bare verdict is

  • intern

    Low-tier generic worker that does exactly what its dispatch prompt states. Use for read-and-judge work whose output is a verdict rather than a change: classifying, triaging findings, checking a claim against a file — even a one-line task. Work that edits code or docs is `employee`: a different test.

  • knowledge-ingestor

    Grooms wiki-memory domains at the moment new evidence arrives — files a scratch/{project}/learned/ capture set onto its pages, dropping what fails admission, and corrects every page the capture reaches, sweeps a whole domain when asked, and runs a sweep proposal the user has approved. Use when a lea

  • principal

    High-tier generic worker that does exactly what its dispatch prompt states. Use only on escalation: `employee` came back wrong or shallow on the same work, or one decision will bind a whole build or epic and can't be cheaply redone. Everything else, security and published work included, is `employee

  • workstream-groomer

    Grooms a workstream's still-open questions and open tasks when new evidence arrives. Use when /pickup, /end-lane or /close-issue dispatches an event groom — even for one row. The full sweep is /groom: a different test.

View source on GitHub

More from radleta/claude-plugins

All 24 plugins →

Development pipeline: brainstorming to idea.md, a lead-driven build with checkpoint reports, one end-of-build review, and the commit workflow

43 skills · 35 commands · 17 agents

0

Wiki-backed persistent memory for Claude Code projects

6 skills · 3 commands · 1 agent

0

Unified code quality and requirements verification with detection categories

3 skills · 1 command · 1 agent

0

Comprehensive commit creation with adaptive analysis and conventional commit format

2 skills · 2 commands · 1 agent

0

Wiki-first codebase investigation plus the capture and ingestion layer that turns what a session learned into durable wiki knowledge

4 skills · 1 agent

0

OWASP Top 10 security verification methodology with detection patterns

3 skills · 1 command · 1 agent

0

Similar plugins

caveman

claude

Talk like caveman. Cut filler. Keep technical details explicit.

22 skills · 1 command · 3 agents · hooks

JuliusBrussee/caveman 119,462

claude-mem

claudecodex

Persistent memory for Claude Code: the next session starts where the last one left off.

29 skills · 1 MCP server · hooks

thedotmack/claude-mem 116,618

Taste Skill

claudecodex

tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.

1 skill

Leonxlnx/taste-skill 114,391

career-ops

claudecodex

Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications

5 skills · hooks

career-ops-hq/career-ops 82,253

mem0

claudecodex

Cross-session memory and token savings for coding agents.

6 skills · 1 agent · 1 MCP server · hooks

mem0ai/mem0 68,893

context7

claudecodex

Up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your coding agent's context.

1 MCP server

upstash/context7 64,709

Maintain scratch-memory?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins24 plugins