prime
Skill3 starsChanged 5 months agoMITView on GitHub
Load project context. Use at the start of every session. Supports quick (default) and full modes.
9 other projects use a version of this.
Adopted by
Most recently taken up by
- malinskibeniamin/skillstheir own version, first seen Oct 17, 2026
- m2ai-portfolio/claude-modstheir own version, first seen Oct 15, 2026
- promptadvisers/claude-codex-workflow-kittheir own version, first seen Oct 8, 2026
- Azure/osdu-spi-stacktheir own version, first seen Oct 2, 2026
- zeuikli/claude-code-workspacetheir own version, first seen Jun 16, 2026
- BriGadja/iapreneurs-claude-code-kittheir own version, first seen May 21, 2026
Used with
Projects with it also use research.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
NVIDIASkillresearch
Before designing a feature, research how comparable libraries solve the problem, check best practices, known pitfalls, and upstream changes. Use when starting a new feature, investigating an unfamiliar area, or the user asks to research something.
Also in 3 projects that use this
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- zeuikli/claude-code-workspace★ 155same name
- promptadvisers/claude-codex-workflow-kit★ 16same name
- BriGadja/iapreneurs-claude-code-kit★ 6same name
- malinskibeniamin/skills★ 4same name
- robzolkos/zolkos-agent-skills★ 3same name
- toneron2/agentic-platform★ 2same name
- jelle/claude-skills★ 1same name
- Azure/osdu-spi-stack★ 0same name
- m2ai-portfolio/claude-mods★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository67 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- optimize-contextSkill
- a11y-checkSkill
- api-contractSkill
- backend-patternsSkill
- build-fixSkill
- checkpointSkill
- clickhouse-ioSkill
- code-review-eccSkill
- code-reviewSkill
- coding-standardsSkill
- config-auditSkill
- continuous-learning-v2Skill
- create-prdSkill
- deps-auditSkill
- e2eSkill
- eval-harnessSkill
- evalSkill
- evolveSkill
- executeSkill
- execution-reportSkill
- extract-sharedSkill
- firecrawlSkill
- frontend-designSkill
- frontend-patternsSkill
- golang-patternsSkill
- golang-testingSkill
- implement-fixSkill
- init-projectSkill
- instinct-exportSkill
- instinct-importSkill
- instinct-statusSkill
- iterative-retrievalSkill
- lazySkill
- learnSkill
- orchestrateSkill
- perf-profileSkill
- plan-eccSkill
- plan-featureSkill
- postgres-patternsSkill
- project-guidelines-exampleSkill
- projectSkill
- prompt-auditSkill
- python-patternsSkill
- python-reviewSkill
- python-testingSkill
- rcaSkill
- refactor-cleanSkill
- releaseSkill
- security-reviewSkill
- seo-kitSkill
- set-working-dirSkill
- skill-createSkill
- strategic-compactSkill
- supadataSkill
- system-reviewSkill
- tddSkill
- tdd-workflowSkill
- test-coverageSkill
- testSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
10 rules. 0 are common, 8 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
8
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 7
Check $ARGUMENTS for mode.
- Very specific: line 89
Read pyproject.toml, package.json, tsconfig.json, composer.json (full - usually short)
- Distinctive: line 95
Test files (just list them)
- Distinctive: line 185
If during planning you realize you need more context, run prime with "full" argument
- Distinctive: line 195
For faster loading: Use prime without arguments (quick mode - 35-40K tokens)
- Distinctive: line 230
Read CLAUDE.md or similar global rules file (FULL)
- Distinctive: line 231
Read README files at project root and major directories (FULL)
- Distinctive: line 232
Read any architecture documentation (FULL)
- Very specific: line 233
Read PRD if it exists (.claude/PRD.md) (FULL)
- Distinctive: line 250
Check current branch and status:
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Check $ARGUMENTS for mode. If "$ARGUMENTS" is "full", follow Full Mode below. Otherwise follow Quick Mode (default).
Quick Mode
Prime Quick: Load Essential Project Context
Objective
Build sufficient understanding of the codebase by analyzing structure, documentation, and key files efficiently. Uses ~35-40K tokens instead of 85K+ (60% savings).
When to Use
- Default choice for most feature planning and development
- Starting a new session
- Getting oriented in a codebase
- Before planning features
When to Use Full Mode Instead
- First time working with a codebase
- Need deep architectural understanding
- Planning major refactors or migrations
- Onboarding to a complex project
Process
1. Analyze Project Structure (Compact)
List tracked files (first 50 only):
git ls-files | head -50Show directory structure (depth 2 max):
find . -maxdepth 2 -type d -not -path '*/\.git/*' -not -path '*/__pycache__/*' -not -path '*/node_modules/*' | sortCheck root directory:
ls -la | head -202. Read Core Documentation (Selective)
ALWAYS READ:
- Project
CLAUDE.md(full - contains project-specific rules and context) requirements.txtorpackage.jsonor similar (full - dependencies)
READ WITH LIMITS:
README.md(limit=100 lines - overview only, skip detailed usage)- Main entry point like
app.py,index.ts,main.py(limit=100 lines - configuration and structure)
SKIP IN QUICK MODE:
- ARCHITECTURE.md (use full mode if needed)
- DEPLOYMENT.md
- Additional documentation files
- PRD (unless specifically needed for the task)
3. Identify Key Files (Efficient)
Use grep/find to identify structure without reading full files:
For Python projects:
# List all modules
ls -la modules/ src/ app/ 2>/dev/null | head -30
# Get function/class signatures (don't read full files)
grep -n "^class \|^def " *.py modules/*.py src/*.py 2>/dev/null | head -40For JavaScript/TypeScript projects:
# List all source files
ls -la src/ lib/ components/ 2>/dev/null | head -30
# Get export signatures
grep -n "^export \|^class \|^function " src/**/*.{js,ts,tsx} 2>/dev/null | head -40Configuration files:
- Read
pyproject.toml,package.json,tsconfig.json,composer.json(full - usually short)
SKIP full reads of:
- Model/schema files (use grep for class names only)
- Service/controller files (use grep for function signatures)
- Template files
- Test files (just list them)
4. Understand Current State (Minimal)
Recent activity (5 commits, not 10):
git log -5 --onelineCurrent branch and status (compact):
git status --short
git branch --show-currentPython version check (if Python project):
python --version 2>&15. Test Infrastructure (Overview Only)
Don't read test files, just list:
ls -la tests/ test/ __tests__/ 2>/dev/null | head -20
find tests/ -name "*.py" -o -name "*.js" -o -name "*.ts" 2>/dev/null | head -10Output Report
Provide a concise summary covering:
Project Overview
- Purpose and type of application
- Primary technologies and frameworks
- Current version/state
Architecture
- Overall structure and organization (based on directory listing)
- Key architectural patterns identified (from CLAUDE.md and file structure)
- Important directories and their purposes
Tech Stack
- Languages and versions
- Frameworks and major libraries (from dependencies file)
- Build tools and package managers
- Testing frameworks (from dependencies)
Core Principles
- Code style and conventions observed (from CLAUDE.md)
- Documentation standards
- Testing approach (from test directory structure)
Current State
- Active branch
- Recent changes or development focus (from last 5 commits)
- Any immediate observations or concerns
Key Files Identified
- Main entry points
- Configuration files
- Module/component structure (high-level only)
Keep the summary concise - use bullet points and clear headers. Focus on actionable information.
Token Budget
Target: 35-40K tokens (vs 85K+ for full mode)
- Structure analysis: ~5-10K
- Core docs: ~10-15K
- File signatures: ~5-10K
- Current state: ~5K
- Report: ~5-10K
Differences from Full Mode
| Aspect | Quick Mode | Full Mode |
| Token usage | 35-40K | 85K+ |
| README | First 100 lines | Full read |
| Source files | grep signatures only | Full reads |
| Test files | List only | Read examples |
| Docs | CLAUDE.md only | All docs |
| Git history | 5 commits | 10 commits |
| File listing | head -50 | All files |
| Use case | 90% of sessions | Deep dives |
Notes
- If during planning you realize you need more context, run prime with "full" argument
- Quick mode captures ~95% of useful context for most tasks
- You can always read specific files later with the Read tool
- Focus is on structure and patterns, not exhaustive coverage
Full Mode
Prime Full: Comprehensive Project Context
Token Usage: 85K+ tokens For faster loading: Use prime without arguments (quick mode - 35-40K tokens)
Objective
Build comprehensive understanding of the codebase by analyzing structure, documentation, and key files in depth.
When to Use
- First time working with a codebase
- Need deep architectural understanding
- Planning major refactors or migrations
- Onboarding to complex projects
- Need to understand all documentation and patterns
Process
1. Analyze Project Structure
List all tracked files:
git ls-filesShow directory structure: On Windows/Linux, run:
# If tree is available
tree -L 3 -I 'node_modules|__pycache__|.git|dist|build|.venv'
# Otherwise use ls/dir
ls -R | grep ":$" | sed -e 's/:$//' -e 's/[^-][^\/]*\//--/g' -e 's/^/ /' -e 's/-/|/'2. Read Core Documentation
- Read CLAUDE.md or similar global rules file (FULL)
- Read README files at project root and major directories (FULL)
- Read any architecture documentation (FULL)
- Read PRD if it exists (.claude/PRD.md) (FULL)
3. Identify Key Files
Based on the structure, identify and read:
- Main entry points (main.py, index.ts, app.py, server.js, etc.) (FULL)
- Core configuration files (pyproject.toml, package.json, tsconfig.json, composer.json) (FULL)
- Key model/schema definitions (FULL)
- Important service or controller files (FULL)
4. Understand Current State
Check recent activity:
git log -10 --onelineCheck current branch and status:
git statusOutput Report
Provide a concise summary covering:
Project Overview
- Purpose and type of application
- Primary technologies and frameworks
- Current version/state
Architecture
- Overall structure and organization
- Key architectural patterns identified
- Important directories and their purposes
Tech Stack
- Languages and versions
- Frameworks and major libraries
- Build tools and package managers
- Testing frameworks
Core Principles
- Code style and conventions observed
- Documentation standards
- Testing approach
Current State
- Active branch
- Recent changes or development focus
- Any immediate observations or concerns
Make this summary easy to scan - use bullet points and clear headers.
Token Budget
Full Mode: 85K+ tokens
Differences from Quick Mode
| Aspect | Quick Mode | Full Mode |
| Token usage | 35-40K | 85K+ |
| README | First 100 lines | Full read |
| Source files | grep signatures only | Full reads |
| Test files | List only | Read examples |
| Docs | CLAUDE.md only | All docs |
| Git history | 5 commits | 10 commits |
| File listing | head -50 | All files |
| Use case | 90% of sessions | Deep dives |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

