agentsclimarketplace

Context

Skill markstent/muster/skills/context

Structured agentic development, shipped as Agent Skills - plan, triage, build, and review with autonomous sub-agents, human approval gates, and automatic review on every diff. Runs in Claude Code, Cursor, Codex, Gemini, Mistral, pi, and any Agent Skills harness.

Install
npx -y skills add markstent/muster --skill context

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 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

(Muster) Write CONTEXT.md - the shared notes every other Muster command reads (your stack, your terms, your past decisions). Run once per repo, then refresh when something major changes or a past decision is reversed. Reads the repo; writes only CONTEXT.md.

SKILL.md

3.5 KB, 807 tokens by cl100k_base, as published. Nobody here has run it

Context

Write and maintain CONTEXT.md at the repo root. It's the shared notes every other Muster command reads - /spec, /triage, and /build - so the agents speak your project's language and respect decisions you've already made. It's working memory for the agents, not a doc for people: keep it dense and current, not pretty.

You read the codebase. You write exactly one file: CONTEXT.md. Touch nothing else.

Step 1 - Detect first run vs refresh

  • If CONTEXT.md does not exist, this is a first run: build it from scratch.
  • If it exists, this is a refresh: read it first, preserve what is still true, and update only what has changed. Never silently drop an existing decision - if you believe one is now wrong, flag it to me before removing it.

Step 2 - Read the repo

  • README.md and any top-level docs.
  • git log --oneline -20 and the most recently changed source files.
  • Package manifests (package.json, pyproject.toml, go.mod, etc.) for the stack and tooling.
  • Any existing docs/adr/, CONTRIBUTING.md, CLAUDE.md/AGENTS.md, or style/standards files - these are existing decisions, not raw material to rewrite.

If something central is ambiguous (the core domain noun, the primary entry point), ask me one targeted question rather than guessing.

Step 3 - Write CONTEXT.md

Use this structure. Scale each section to the repo; omit a heading only if it genuinely does not apply.

# CONTEXT

## Stack
[Languages, frameworks, key libraries, test runner, package manager.
 One line each.]
Test command: [single command that runs the full suite, e.g. `pytest -q`.
 /build re-runs exactly this to gate every task; /triage requires it before a
 task is agent-ready. Keep it one runnable line.]

## Domain glossary
[The nouns and verbs of this project, defined. The vocabulary every spec,
 task, and review must use. Two columns: term - meaning.]

## Architecture
[The shape of the system in a few sentences. Entry points, major modules,
 how data flows. Name the directories that matter.]

## Conventions
[How code is written here: naming, error handling, test style, anything a
 new contributor would get wrong. Point to the enforcing config where one
 exists rather than restating it.]

## Decisions (ADRs)
[Dated, one-line architectural decisions and their rationale. Newest first.
 Format: YYYY-MM-DD - decision - why. These are binding on /triage and /build.]

## Out of scope
[Things deliberately not built or not done, and why. Mirrors anything in
 .out-of-scope/.]

Step 4 - Summarise

Print what you wrote or changed:

## Context - ✅ CONTEXT.md [created | updated]

Saved your project notes - the shared memory every Muster command now reads.

| | |
|---|---|
| Stack | [one line] |
| Glossary | [N terms] |
| Decisions | [N recorded] |
| Changed | [what this refresh added or revised, or "first run"] |

**Next:** run /spec to turn a resolved idea into issues.

Rules

  • Write only CONTEXT.md. Never modify code, branches, or other files.
  • On a refresh, preserve existing decisions unless I approve removing one.
  • Keep it terse. This is memory for agents, not a tutorial.
  • Record decisions as they are made - a reversed decision is an update here, not a deletion of history.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,984. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.