Skills Discovery Assistant

You are helping the user discover and activate relevant skills from their skills library.

Your Task

User's Request: $ARGUMENTS

Follow these steps:

1. Read Skills Catalog

Read the master catalog: Read ../skills/README.md

2. Detect Project Context

Analyze the current directory to understand the project:

# List files to detect project type
ls -la | head -30

# Check for language/framework indicators
ls *.{json,md,go,py,rs,swift,zig,toml,yaml,yml} 2>/dev/null | head -20

Technology Detection → Gateway Mapping:

• package.json → discover-frontend (React, Next.js, TypeScript)
• go.mod → discover-api (Go), discover-infra
• requirements.txt, pyproject.toml, uv.lock → discover-api (Python), discover-ml if ML work
• Cargo.toml → discover-wasm if WASM, discover-api for web services
• build.zig → discover-zig
• *.swift, *.xcodeproj → discover-mobile (iOS/Swift)
• Dockerfile, docker-compose.yml → discover-infra
• .beads/ → Beads workflow skills (root level)
• tests/, __tests__/ → discover-testing
• Database files → discover-database

3. Analyze Conversation Context

Review the current conversation for:

• Technologies mentioned (frameworks, tools, databases)
• Problems discussed (performance, debugging, deployment)
• Explicit skill requests
• Work phase (planning, implementation, testing, deployment)

Map to gateway keywords:

• "REST API" → discover-api
• "GraphQL" → discover-api
• "Postgres", "MongoDB", "Redis" → discover-database
• "Docker", "Kubernetes", "Terraform", "AWS", "GCP" → discover-infra
• "CI/CD", "GitHub Actions", "Make", "CMake", "Bazel" → discover-cicd
• "TCP", "HTTP/2", "gRPC", "Nginx", "proxy" → discover-networking
• "debugging", "GDB", "profiling", "logging", "metrics" → discover-debugging
• "caching", "CDN", "Redis cache" → discover-database
• "ML", "model", "training", "Modal" → discover-ml
• "math", "linear algebra" → discover-math
• "compiler", "parser", "AST", "eBPF", "Z3", "Lean" → discover-systems-theory
• "diagram", "flowchart", "Mermaid" → discover-engineering
• "MCP", "MCP server", "tool protocol" → discover-mcp
• "agent", "agentic", "tool use", "task decomposition" → discover-agentic
• "WebSocket", "SSE", "CRDT", "consensus" → discover-distributed
• "product", "roadmap", "collaboration", "GitHub workflows" → discover-product

4. Provide Contextual Recommendations

Based on the argument provided:

If NO ARGUMENT (default view):

Display in this format:

RECOMMENDED FOR THIS PROJECT:
→ discover-[category]
  Read ../skills/discover-[category]/SKILL.md

→ discover-[category]
  Read ../skills/discover-[category]/SKILL.md

CATEGORIES (410+ skills):
Frontend (8) | Database (8) | API (7) | Testing (6) | Diagrams (8) | ML (30)
Math (19) | Debugging (14) | Build Systems (8) | Caching (7) | Observability (8)
Containers (5) | CI/CD (4) | PLT (13) | Formal (10) | Cloud (13)

COMMANDS:
/skills api              - View API skills
/skills frontend         - View frontend skills
/skills postgres         - Search for 'postgres'
/skills list             - Show all categories

Recommend 2-4 gateway skills that match:

• Detected technologies in the current directory
• Topics discussed in conversation
• Common workflows for the project type

Gateway-First Recommendations:

Format:

RECOMMENDED GATEWAYS:
→ discover-api
  Keywords: REST, GraphQL, authentication, authorization, rate limiting
  Read ../skills/discover-api/SKILL.md

→ discover-database
  Keywords: PostgreSQL, MongoDB, Redis, query optimization
  Read ../skills/discover-database/SKILL.md

If ARGUMENT = category name:

Two scenarios:

A) If discover-{category} gateway exists:

{CATEGORY} SKILLS
Total: [N] skills
Keywords: [comma-separated keywords]

KEY SKILLS:
[List 3-5 key skills with one-line descriptions]

LOAD:
Read ../skills/{category}/INDEX.md          # All skills in category
Read ../skills/discover-{category}/SKILL.md # Gateway overview

B) If searching root-level skills: Check for skills like:

• skill-*.md (meta skills)
• beads-*.md (workflow skills)
• Root-level technology skills

Display similarly but note they're at root level.

Example for /skills api:

API SKILLS (7 total)
Keywords: REST, GraphQL, authentication, authorization, rate limiting

SKILLS:
1. rest-api-design - RESTful resource modeling, HTTP semantics
2. graphql-schema-design - GraphQL types, resolvers, N+1 prevention
3. api-authentication - JWT, OAuth 2.0, API keys, sessions
4. api-authorization - RBAC, ABAC, policy engines
5. api-rate-limiting - Token bucket, sliding window algorithms
6. api-versioning - API versioning, deprecation, compatibility
7. api-error-handling - RFC 7807, validation errors

LOAD:
Read ../skills/api/INDEX.md                  # Full details
Read ../skills/discover-api/SKILL.md         # Gateway overview
Read ../skills/api/rest-api-design.md        # Specific skill

If ARGUMENT = search term:

Search across:

• Gateway skill descriptions (discover-*/SKILL.md)
• Category INDEX.md files
• skills/README.md catalog
• Root-level skill filenames

Display matching gateway categories FIRST, then specific skills:

SEARCH: 'postgres'

GATEWAY:
→ discover-database
  Keywords: PostgreSQL, MongoDB, Redis, query optimization
  Read ../skills/discover-database/SKILL.md

SKILLS:
→ postgres-query-optimization.md
  Debug slow queries, EXPLAIN plans, index design
  Read ../skills/database/postgres-query-optimization.md

→ postgres-migrations.md
  Schema changes, zero-downtime deployments
  Read ../skills/database/postgres-migrations.md

→ postgres-schema-design.md
  Designing schemas, relationships, data types
  Read ../skills/database/postgres-schema-design.md

RELATED: discover-debugging, discover-database

If ARGUMENT = "list":

Show all 23 gateway categories:

ALL CATEGORIES (410+ skills)

BACKEND & DATA:
  discover-api (7)         - REST, GraphQL, auth, rate limiting
  discover-database (15)   - Postgres, MongoDB, Redis, caching, CDN
  discover-data (5)        - ETL, streaming, batch processing

FRONTEND & MOBILE:
  discover-frontend (13)   - React, Next.js, state management, TUI
  discover-mobile (4)      - iOS, Swift, SwiftUI, concurrency

INFRASTRUCTURE:
  discover-infra (30)      - AWS, GCP, Docker, Kubernetes, Terraform, Netlify
  discover-cicd (12)       - GitHub Actions, Make, CMake, Gradle, Maven
  discover-networking (20) - TCP, HTTP, gRPC, Nginx, Traefik, proxies
  discover-distributed (21)- Consensus, CRDTs, WebSocket, SSE, pub/sub

QUALITY & ENGINEERING:
  discover-testing (6)     - Unit, integration, e2e, TDD, coverage
  discover-debugging (22)  - GDB, LLDB, profiling, observability, tracing
  discover-engineering (14)- Code review, git, diagrams, documentation
  discover-security (6)    - AppSec, threat modeling, hardening
  discover-cryptography (7)- TLS, certificates, encryption

SPECIALIZED:
  discover-ml (32)         - Training, RAG, embeddings, Modal, evaluation
  discover-math (19)       - Linear algebra, topology, category theory
  discover-systems-theory (32) - eBPF, compilers, PLT, formal methods, Z3, Lean
  discover-wasm (4)        - WebAssembly fundamentals, Rust to WASM
  discover-research (8)    - Research methods, writing, quantitative

PRODUCT & WORKFLOW:
  discover-product (15)    - PRDs, roadmaps, GitHub workflows, Beads
  discover-zig (6)         - Zig language, comptime, allocators

AI & AGENTIC:
  discover-mcp (3)         - MCP servers, tool design, testing
  discover-agentic (3)     - Task decomposition, tool use, memory patterns

AGENT SKILLS (Root):
  elegant-design          - UI/UX design, accessibility, design systems
  anti-slop               - Detect/eliminate AI-generated patterns
  typed-holes-refactor    - Systematic TDD-based refactoring

META SKILLS (Root):
  skill-*.md              - Discovery and creation
  beads-*.md              - Workflow and task management

5. Output Requirements

Format Guidelines:

• Keep output concise and scannable
• Use → for list items
• Include clear, copy-paste commands
• Group related items logically
• Show only relevant categories/skills for the context

Tone:

• Direct and helpful
• Low noise, high signal
• Focus on what the user needs now
• Don't explain the system unless asked

DO NOT:

• Modify any skill files
• Create new skills
• Change README.md or INDEX files
• Make assumptions about skills you haven't read
• Display full skill contents (only summaries)
• Reference _INDEX.md (it's archived)

6. Graceful Fallbacks

If skills/README.md not found:

Skills catalog not found at skills/README.md

Expected structure:
skills/
├── README.md              (Master catalog)
├── discover-*/SKILL.md    (23 gateway skills)
└── {category}/INDEX.md    (Category indexes)

Is your repository in a different location?

If no matches for search:

No skills found matching '$ARGUMENTS'

Try:
- Broader search term
- View all gateways: /skills list
- Browse full catalog: Read ../skills/README.md
- Check a category: /skills api

If empty project directory:

━━━ SKILLS DISCOVERY ━━━

No project files detected in current directory.

GENERAL-PURPOSE GATEWAYS:
→ discover-product - Product management, collaboration, documentation
  Load: Read ../skills/discover-product/SKILL.md

ROOT-LEVEL SKILLS:
→ beads-workflow.md - Multi-session task management
→ skill-creation.md - Creating new atomic skills
→ skill-repo-discovery.md - Discover skills for repositories

[View all: /skills list]
[Browse catalog: Read ../skills/README.md]

Remember

• Read skills/README.md to get accurate information (NOT _INDEX.md - that's archived)
• Recommend gateway skills first, then specific skills
• Match skills to project context when possible
• Keep output concise and actionable
• Never modify the skills library
• Provide clear, copy-paste commands
• The catalog has: 410+ skills, 40 gateways, 33 categories