Claude Night Market

A plugin marketplace for Claude Code. Install only the plugins you need to run git workflows, code review, spec-driven development, and autonomous agents from inside your Claude Code session.

Install

Requires Claude Code 2.1.16+ and Python 3.9+ for hooks.

# Add the marketplace, then install the plugins you want
/plugin marketplace add athola/claude-night-market
/plugin install sanctum@claude-night-market    # Git workflows
/plugin install pensive@claude-night-market    # Code review
/plugin install spec-kit@claude-night-market   # Spec-driven dev

Run claude --init once after installing. Prefer one command? npx skills add athola/claude-night-market installs everything; opkg i gh@athola/claude-night-market --plugins sanctum,pensive installs a subset. Full options are in the Installation Guide.

If the Skill tool is unavailable, read skill files directly at plugins/{plugin}/skills/{skill-name}/SKILL.md.

Everyday Use

Night Market is built around the loop you already work in. A typical feature runs end to end on a handful of commands:

1. Start a feature. /attune:mission routes you through brainstorm, specify, plan, and execute phases.
2. Write the code. imbue enforces a failing test first, so implementation follows the test, not the other way around.
3. Review before you push. /full-review runs a multi-discipline pass; /refine-code cleans up duplication and dead code.
4. Ship it. /prepare-pr runs quality gates and leaves a clean git state ready for a pull request.
5. Pick up where you left off. /catchup rebuilds context from recent git history after a break.

The commands you reach for most:

Full task-by-task walkthroughs are in the [Common Workflows Guide][workflows].

What's Inside

23 plugins in four layers. Each installs independently, and dependencies pull their shared runtime automatically.

Foundation is the base every other layer builds on: leyline (auth, quotas, error patterns, trust verification), sanctum (git, commits, PR prep, sessions), and imbue (TDD enforcement, proof-of-work, scope guarding).

Utility handles cross-cutting concerns: conserve (context and token optimization), conjure (delegation to Gemini and Qwen), hookify (a behavioral rules engine with a security catalog), egregore (autonomous agent orchestration), herald (notifications), and oracle (local ML inference).

Domain is where the day-to-day work happens: pensive (code and architecture review), attune (project lifecycle), spec-kit (spec-driven development), parseltongue (Python), minister (GitHub issues and DORA metrics), memory-palace (knowledge organization), archetypes (architecture paradigms), gauntlet (codebase learning), phantom (computer use), scribe (documentation and slop detection), scry (recordings), tome (research), and cartograph (codebase visualization).

Meta improves the system itself: abstract (skill authoring, hook development, evaluation, and skill-stability tracking).

The full skill, command, and agent inventory is in the Capabilities Reference.

Safety

⚠️ Plugins run inside your Claude Code session and can read or edit your repo, run shell commands, and call external services. Review any plugin before installing it.

Three guards reduce the blast radius, but none replace your own review:

• TDD gates (imbue) block implementation writes that lack a failing test.
• Destructive-command blockers (conserve, hookify) auto-approve safe commands and halt or warn on rm -rf, git push --force, and production-shaped targets.
• Additive-bias audits (leyline) flag unjustified additions before commit.

CONSTITUTION.md holds the immutable rules that override any conflicting skill or hook; STEWARDSHIP.md is the maintenance contract.

Network and data

Two hooks reach the network using your existing GitHub credentials. Both fail silently when gh is unauthenticated or the network is unavailable, and both can be turned off.

• Star prompt (leyline, plugins/leyline/hooks/auto-star-repo.sh). On session start it checks whether you have starred athola/claude-night-market, using your gh CLI auth or a GITHUB_TOKEN / GH_TOKEN env var. It only reads star status and asks once per session; it never stars or unstars without your consent. Opt out by setting CLAUDE_NIGHT_MARKET_NO_STAR_PROMPT=1.
• Learnings and insights posting (abstract, plugins/abstract/hooks/post_learnings_stop.py). On session stop, if ~/.claude/skills/LEARNINGS.md has content, it posts a skill-usage summary (and may promote high-severity items to issues) via your authenticated gh CLI. The target is detected at runtime: a target_repo override in ~/.claude/skills/discussions/config.json, otherwise the current repo from gh repo view. Posting defaults to on; opt out by setting auto_post_learnings to false in that config file.

Requirements

• Claude Code 2.1.16+ (2.1.32+ for agent teams, 2.1.38+ for security features).
• Python 3.9+ for hooks (macOS ships 3.9.6). Hook code must stay 3.9-compatible; plugin packages may target 3.10+ via virtual environments. See the [Plugin Development Guide][dev-guide] for the rules.

What's New

1.9.16 adds Domain-Driven Design as the fourteenth architecture paradigm, along with the review lens that enforces it. The archetypes:architecture-paradigm-domain-driven skill treats DDD as modeling a business in its own language, and treats layering, mapping, DTOs, and command objects as separable machinery a domain model may or may not need. Its companion ceremony-audit module in pensive:architecture-review reviews for passthrough mappers, twin types, and speculative DTOs, with an IO-boundary counter-signal so it does not flag a boundary mapper doing real work.

The release also adds two pensive review capabilities as modules inside existing skills. The rust-review skill gains a concurrency-patterns rule that flags hand-rolled multi-task orchestration (several tokio::spawn handles torn down with consecutive abort() calls) that a single select! loop expresses more safely. The performance-review skill gains three manual memory-allocation lenses: unbounded collections from an external source, hot-path recompute that should be memoized, and serial blocking I/O over an unbounded set. These are lenses a reviewer applies by reading the code, with no AST automation behind them.

On the infrastructure side, cartograph's 40 tests now run in a gate that previously reported them as a clean skip, and the typecheck gate resolves mypy from each plugin's own environment so it means the same thing locally and in CI. See the CHANGELOG for the full history.

Plugin Development

make create-plugin NAME=my-plugin
make validate
make lint && make test

A plugin directory holds .claude-plugin/plugin.json (metadata) plus any of commands/, skills/, hooks/, agents/, and tests/, with a Makefile and pyproject.toml. See the [Plugin Development Guide][dev-guide] for structure and naming conventions.

Documentation

• Installation Guide
• Quick Start
• [Common Workflows][workflows]
• [Plugin Development Guide][dev-guide]
• Capabilities Reference
• Tutorials
• Architecture Decision Records

Per-plugin pages are in book/src/plugins/.

Stewardship and Contributing

Every plugin is entrusted to the community: steward rather than own, and think several iterations ahead. Each plugin maintains its own tests and docs; run make test at the repo root to execute every suite, and /stewardship-health to view per-plugin health. Contribution guidelines live in the [Plugin Development Guide][dev-guide].

Acknowledgements

Night Market builds on Anthropic Claude Code and integrates with [github/spec-kit][spec-kit-upstream] (v0.5.0), [obra/superpowers][superpowers-upstream] (v5.0.7, see the [integration guide][superpowers-doc]), and three patterns adapted from [QAInsights/Quillx][quillx]. Per-plugin attributions are in each plugin's pyproject.toml.

License

MIT

[dev-guide]: