Rapidly acquire a mental model of any unfamiliar codebase — from a 500-line script to a 100M+ line monorepo. This skill transforms raw code into structured intelligence: architecture maps, entry points, data flows, security surfaces, and onboarding confidence scores.
Rapidly acquire a mental model of any unfamiliar codebase — from a 500-line script to a 100M+ line monorepo. This skill transforms raw code into structured intelligence: architecture maps, entry points, data flows, security surfaces, and onboarding confidence scores.
This skill transforms raw code into structured intelligence: architecture maps, entry points, data flows, security surfaces, and onboarding confidence scores.
Domain: knowledge
Skill Identity
Attribute
Value
Domain
Knowledge Operations
Skill ID
codebase-onboarding
Version
1.0.0
Hacker Laws
Law 1 (Know Your Battlefield), Law 3 (Intelligence Over Force), Law 9 (Systematic Over Random)
Related Skills
knowledge-ops, deep-research, social-intelligence
Purpose
Rapidly acquire a mental model of any unfamiliar codebase — from a 500-line script to a 100M+ line monorepo. This skill transforms raw code into structured intelligence: architecture maps, entry points, data flows, security surfaces, and onboarding confidence scores.
In security contexts, codebase onboarding is the first step before code audits, vulnerability research, exploit development, and supply chain analysis.
Scope Modes
Three modes based on what you need to know, not how much time you have:
Mode
When to Use
Output
Targeted
You know what you're looking for (e.g., "find all auth code")
Focused map of specific subsystem
Exploratory
You need to understand a feature area or module
Module-level architecture + data flows
Comprehensive
Full audit, exploit research, or security review
Complete intelligence package
Choose mode before starting. Switching modes mid-session is allowed but requires re-baselining.
Phase 0: Search-First (All Modes)
Before reading any code:
Find existing documentation
README, CONTRIBUTING, ARCHITECTURE, docs/, wiki/
OpenAPI/Swagger specs, Protobuf definitions
CI/CD config (reveals build structure and test commands)
Kotlin, Scala, Swift, Objective-C, Erlang, Elixir, Haskell, COBOL, and other languages require primarily manual analysis. Use Phase 0 docs-first approach and lean on test files for behavior discovery.
100M+ Line Strategy
For very large codebases (100M+ LOC):
Index First: Run ctags/cscope before reading any files
Smart Sampling: Focus on files with highest churn (git log), most imports, or security-critical paths
Divide & Conquer: Treat each top-level module as a separate Targeted-mode session
Boundary Focus: Understand module interfaces (APIs, contracts) before internals
Avoid Full Reads: Never attempt to read entire large files; sample entry, middle, and exit sections
Output Format
Confidence Score
Rate onboarding completeness per area:
Score
Meaning
0–20
Uncharted — no meaningful understanding
21–40
Partial — know structure, not behavior
41–60
Functional — can navigate, some gaps
61–80
Solid — understand core flows and surfaces
81–100
Expert — deep understanding, audit-ready
Report as: Overall: 72/100 | Auth: 85 | Data Layer: 60 | API Surface: 78 | Internal Logic: 65
Structured Intelligence Package
{"project":"target-name","analyzed_at":"2026-05-11","mode":"Comprehensive","language_primary":"Go","framework":"Gin + GORM","loc_total":85000,"confidence":{"overall":72,"auth":85,"data_layer":60,"api_surface":78,"internal_logic":65},"entry_points":["cmd/server/main.go","cmd/worker/main.go"],"architecture":"Monolith with event-driven background workers","security_surfaces":{"auth":"JWT via middleware/auth.go","input_validation":"Partial — missing in admin routes","secrets":"env vars via config/config.go","dangerous_patterns":["SQL concatenation in reports/query.go:145"]},"gaps":["Payment flow not traced","gRPC service definitions not reviewed"],"next_steps":["Audit reports/query.go for SQLi","Review payment/ module"]}
