claude

compost

One software-engineering workflow composted from Matt Pocock's skills, superpowers, pstack, Imbue's blueprint, and autonomous-sdlc: specs with BDD acceptance criteria, tracer-bullet issues in your tracker, builds that fit tests into the existing suite, automatic subagent review, browser diff review, semantic lint for Python, a prune skill that keeps docs to what the code can't say, and a turn skill that works in upstream changes.

7 on GitHubby Joshua Oliphantv0.33.2MITupdated todaydevelopment
/

Plugin compost from marketplace oliphant-plugins in JoshuaOliphant/claude-plugins. Commands read from the repo's manifest — today.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (18)

  • deepen

    Designs deep modules, meaning small interfaces with a lot of behaviour behind them, seams where something really varies, and tests at the interface. Use when the user says "design this interface", "where should the seam go", "refactor this for testability", "this module is hard to test", "find deepe

  • deeper-ask

    Makes a request deeper, not bigger, by asking what a top practitioner would do, the rawest material, the methods and tools, and what independent check would show the result right, then rewriting the request to ask for exactly that. Use when the user says "deeper ask", "make this deeper, not bigger",

  • diagnose

    Finds a bug's root cause by first building a fast, deterministic check that goes red on it, then testing ranked hypotheses one variable at a time. Use before proposing any fix, when something is broken, throwing, failing, flaky, slow, or regressed - "debug this", "diagnose", "why is this failing", "

  • fan-out

    Settles a contested design or an unclear cause by running several independent attempts in parallel, judging them against each other on a rubric, and synthesizing the winner. Use when one attempt would lock in the wrong shape or anchor on the first theory - "try a few approaches", "compare designs",

  • finish

    Integrates reviewed work by testing it against the branch it will merge into, opening or updating the pull request, and merging or discarding only on the user's word. Use when the user says "finish this", "wrap it up", "open the PR", "update the PR", "ready to merge", "merge it", "clean up the workt

  • implement

    Drives filed issues to a reviewed, draft-PR-ready branch, running build, verify, and review for each issue on its own and working independent issues in parallel. Use once issues exist - "implement this", "implement the spec", "work the issues", "build the tickets", "start on #42", "keep going until

  • pause

    Stops work at a safe point with everything committed and a resumable status comment on the parent issue, and picks paused work back up from the tracker and git. Use when the user says "pause", "stop here for today", "wrap up for now", "hand this off", "write a handoff", "pick up where we left off",

  • prune

    Finds doc drift, where documentation says something the code no longer does, and settles it in the code's favour. What a doc says that the code could say moves into names, types, tests, or a comment at the point of use; the doc then goes, leaving only the core docs (README, AGENTS.md, CONTEXT.md, AD

  • retro

    Looks back over a coding session and files the changes to the agent's environment (checks, pointers, instruction files, tooling, access) that would have made it go better. Use when the user says "retro", "run a retrospective", "what slowed you down", or "what should we change so this goes smoother n

  • review

    Reviews a verified change with subagents on two axes, Standards and Spec, then works through the findings that survive. Run it after compost:verify passes on every issue, without being asked. Use when someone says "review this branch", "review my changes", "check this against the issue", "does the a

  • setup

    Configures a repo once for the compost skills, recording its issue tracker, domain doc layout, core docs, test convention, verify gates, and observability stack under docs/agents/, and offers to move CLAUDE.md into AGENTS.md. Use when the user says "set up compost", "configure this repo", "set up th

  • skill-drift

    Reviews the skill-drift log that the skill-drift mod writes, and turns skills that keep drifting across sessions into draft fixes. Use when asked "review skill drift", "which skills keep going wrong", or on the daily scheduled run.

  • slice

    Breaks an agreed spec into tracer-bullet issues in the repo's tracker, each with its acceptance criteria, files, interfaces, and blocking edges. Use when the user says "slice this", "break this into issues", "make tickets", "create the issues", "make a plan", "write the implementation plan", "plan t

  • spec

    Turns a request for new behavior, a fuzzy idea, or a rough doc into an agreed spec of user stories with Given/When/Then acceptance criteria, posted as the parent issue, before any code is written. Use when the user asks in plain words for a feature or a change in behavior that no issue or spec pins

  • thin

    Finds tests that no longer earn their place, where every bug they would catch another test already catches, and folds or deletes them on the user's word. Per-test coverage narrows the candidates, a full mutation kill matrix judges them, and a hand-made break confirms each one before it goes. Use whe

  • turn

    Checks the skill repos compost was made from for upstream changes worth working in, and adds skill repos to watch as references. Use when asked "anything upstream worth pulling in?", "turn the pile", "check superpowers / Matt Pocock / pstack for updates", "add this skill repo as a reference", or on

  • verify

    Proves a change works from fresh evidence before anyone says it does. Use before claiming done, fixed, or passing, before a commit or PR, and when the user asks "is this ready", "verify this", "prove it works", "are the acceptance criteria met", "check coverage", "run mutation testing", "would the t

  • wait-what

    Re-pitches whatever the user stopped following, with the premise they missed, in plain words and the glossary's terms. Use when the user types /compost:wait-what; only they know when they have lost the thread.

Subagents (1)

  • reviewer

    Reviews a branch diff on one axis, Standards or Spec, or on both for a small diff, and returns findings backed by file:line evidence. Used by the /compost:review-changes workflow; delegate to it directly when a diff needs an independent read against the repo's documented standards or an issue's acce

Hooks (1)

  • Stop

    uv run --quiet "${CLAUDE_PLUGIN_ROOT}/scripts/stop_guard.py"

View source on GitHub

More from JoshuaOliphant/claude-plugins

All 8 plugins →

Capture solved problems, retrieve past solutions, and graduate accumulated knowledge into living context. Skills: compound-capture, compound-retrieve, compound-graduate.

3 skills · 1 command · 1 agent

7

Turn a Claude project or a folder into a learning harness for one subject. Lesson sessions probe your edge, plan a minimal dependency tree, and teach one checked reasoning step at a time; Practice sessions train retrieval with a hint ladder; Field sessions grade an artifact you produce against a pre-committed rubric. A knowledge graph, an append-only ledger and a frontier record carry it across sessions. Adapted from learnharness.

2 skills

7

Install a local, Docker-free OTLP observability stack (OpenTelemetry → Vector → JSONL, optionally VictoriaLogs + VictoriaMetrics) into a project and instrument the code, with a scan-and-propose step for domain-appropriate spans and metrics — then query that telemetry as a development feedback signal (observability-query skill, status --json contract, scripted end-to-end verify).

2 skills

7

superfan

claude

Superfans: one headless Claude per component, obsessively watching its telemetry through your read-only MCP servers (Splunk, Prometheus, Kubernetes, ...) on a local schedule. Findings land in a local inbox, page-worthy ones raise a clickable desktop notification that opens a triage session, and every verdict feeds a ledger that teaches the superfan what normal looks like.

2 skills

7

Process information for real understanding with the explain-back skill: explain from memory, grade against the real source, struggle-then-teach per gap, output Mochi cards and a resumable session record. An antidote to the illusion of clarity.

1 skill · 1 MCP server

7

Build web apps where an AI agent dynamically generates HTML UI using hexagonal architecture with HTMX

1 skill

7

Similar plugins

superpowers

claudecodex

Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques

15 skills · hooks

obra/superpowers 316,836

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

Build powerful, stateful CLI interfaces for any GUI application using the cli-anything harness methodology.

5 commands

HKUDS/CLI-Anything 58,665

Automates browser interactions for web testing, form filling, screenshots, and data extraction

12 skills

vercel-labs/agent-browser 50,373

Use Alibaba Open Code Review from local Codex.

2 skills

alibaba/open-code-review 50,068

slidev

claude

Claude Code skill for creating and presenting Slidev slidedecks

1 skill

slidevjs/slidev 49,452

Maintain compost?

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 plugins8 plugins