Night Market Skill — ported from claude-night-market/conserve. For the full experience with agents, hooks, and commands, install the Claude Code plugin.

Context Map#

Generate a compressed context map for the current project. The map pre-compiles structural knowledge that AI assistants would otherwise discover through expensive Read/Grep calls, saving thousands of tokens per session.

When to Use#

• At the start of a session to understand project layout
• Before implementing features to identify entry points
• When exploring an unfamiliar codebase
• To reduce token waste from Read calls
• To identify hot files (high blast radius) before changes

What It Detects#

CategoryDescriptionStructureDirectory layout with file counts and languagesDependenciesMulti-ecosystem: Python, Node, Rust, Go, JavaFrameworksFramework detection from dependency analysisEntry Pointsmain.py, index.ts, CLI scripts, etc.Import GraphFile-to-file import relationshipsHot FilesFiles imported by 3+ others (high blast radius)RoutesFastAPI, Flask, Express, Hono API endpointsEnv VarsEnvironment variable references with defaultsMiddlewareAuth, CORS, rate-limit, logging patternsModels/SchemasSQLAlchemy, Django, Pydantic, Prisma definitionsToken SavingsEstimated tokens saved vs manual exploration

Procedure#

1. Run the scanner on the project root:

bash

PYTHONPATH="$(find . -path '*/conserve/scripts' -type d \
  -print -quit 2>/dev/null || \
  echo 'plugins/conserve/scripts')" \
  python3 -m context_scanner .

1. Present the output to the user as the project overview.

2. Use the context map to guide subsequent file reads. Prioritize hot files and entry points first.

Options#

Output#

• --format json for structured output
• --max-tokens N to adjust output size (default: 5000)
• --output FILE to save to a file

Modes#

• --blast FILE to show blast radius for a specific file
• --section NAME to output a single section (routes, deps, env, hot-files, models, structure, middleware, frameworks)
• --wiki-only to generate wiki articles without stdout

Opt-out#

• --no-cache to force a fresh scan
• --no-wiki to skip wiki article generation

Wiki Articles#

The scanner generates per-topic knowledge articles in .codesight/ for selective context loading:

bash

python3 scanner.py .
# Creates .codesight/INDEX.md, auth.md, database.md, etc.

Load only what you need per session instead of the full map:

bash

python3 scanner.py --section routes .
# ~200 tokens vs ~5,000 for the full map

Example Output#

text

# Context Map: myproject
Files: 127

## Structure
  src                  42 files (Python)
  tests                18 files (Python)
  docs                  5 files (Markdown)

## Dependencies (Python)
Package manager: uv
  - fastapi 0.104.0
  - pydantic 2.5.0
  - sqlalchemy 2.0.0
  ...12 more

## Frameworks Detected
  - FastAPI
  - SQLAlchemy
  - Pytest

## Routes
  GET    /users          (src/routes/users.py)
  POST   /users          (src/routes/users.py)
  GET    /users/{id}     (src/routes/users.py)

## Hot Files (high blast radius)
  - src/models/base.py (12 importers)
  - src/utils/auth.py (8 importers)

## Environment Variables
  - DATABASE_URL (required)
  - SECRET_KEY (has default)

## Token Savings: ~12,600 tokens saved
  Routes: ~1,200
  Hot files: ~300
  Env vars: ~200
  File scanning: ~10,200