Claude-skill-registry dev-context

Context architecture skill for aligning work. Use when starting new features, making architectural decisions, or setting up implementation work. Generates interconnected ADR/Design/Spec/Plan scaffolding that agents can navigate and humans can enrich.

install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/dev-context" ~/.claude/skills/majiayu000-claude-skill-registry-dev-context && rm -rf "$T"
manifest: skills/data/dev-context/SKILL.md
source content

Dev Context (Lean)

Set up working context for GitHub issues. Create formal ADRs when needed.

Two Modes

Mode 1: Issue Working Space (Default)

Trigger: "Set up context for issue #115", "Create working space for feature X"

Creates:

docs/issues/{num}-{short-name}/
├── README.md   # Status, GH link, quick reference
├── plan.md     # Implementation approach
└── notes.md    # Working notes (starts empty)

Process:

  1. Create folder
    docs/issues/{num}-{short-name}/
  2. Generate README.md with issue link and status
  3. Generate plan.md with implementation approach
  4. Create empty notes.md for working notes
  5. Return file paths

README.md template:

# Issue #{num}: {Title}

**GitHub**: {link}
**Status**: In Progress
**Created**: {date}

## Overview
{Brief description from issue}

## Related
- Plan: [plan.md](./plan.md)
- Notes: [notes.md](./notes.md)

plan.md template:

# Implementation Plan: {Title}

## Approach
{High-level approach}

## Tasks
- [ ] {Task 1}
- [ ] {Task 2}
- [ ] {Task 3}

## Files to Modify
- {file 1}
- {file 2}

## Notes
{Implementation notes as work progresses}

Mode 2: Architecture Decision Record (ADR)

Trigger: "Create ADR for choosing X over Y", "Document the decision about Z"

Creates:

docs/adr/{NNN}-{decision-name}.md

Process:

  1. Find next ADR number (glob
    docs/adr/*.md
    , or start at 001)
  2. Create
    docs/adr/
    directory if needed
  3. Generate ADR from template in
    references/templates/adr.md
  4. Return file path

When to use ADRs:

  • Major architectural decisions (database choice, framework selection)
  • Decisions that will be questioned later ("why did we do it this way?")
  • Decisions with significant tradeoffs worth documenting
  • NOT for routine implementation choices

Note: Most decisions don't need formal ADRs. They get captured in devlogs and synthesized into

docs/moc/decisions.md
during periodic maintenance. Use ADRs sparingly for decisions that truly warrant formal documentation.


What This Skill Does NOT Do

This lean version intentionally omits:

Old FeatureWhy RemovedAlternative
PRD generationGitHub issue IS the PRDWrite clear issues
Design docs per featureOnly major systems need design docs
docs/design/
for systems
Spec docsPlan IS the specUse plan.md
Cross-reference ceremonyOverkill for most workKeep it simple

Quick Reference

RequestModeOutput
"Set up context for issue #115"Issue
docs/issues/115-*/
"Create working space for auth"Issue
docs/issues/{num}-auth/
"Create ADR for X vs Y decision"ADR
docs/adr/NNN-*.md
"Document decision about caching"ADR
docs/adr/NNN-*.md

Lifecycle

Issue Working Space

  1. Create when starting multi-session work
  2. Update plan.md and notes.md as you work
  3. Extract patterns to
    ai_docs/
    if reusable
  4. Delete folder when issue closes

ADRs

  1. Create for major decisions
  2. Never edit once accepted (write new ADR if decision changes)
  3. Reference from
    docs/moc/decisions.md
    for discoverability

Part of the dev-* Family

SkillPurpose
dev-context
Set up working space, create ADRs
dev-explore
Generate/update MOC documentation
dev-reports
Write devlogs and status reports
dev-inquiry
Technical investigation and spikes

See Also

  • docs/issues/README.md
    - Working space conventions
  • ai_docs/maintenance-workflows.md
    - When to extract patterns
  • docs/moc/decisions.md
    - Synthesized decision history