skillshop-ostyles/agents-understanding
Codebase understanding skills: architecture, onboarding, knowledge preservation.
API surface documenter: extracts ALL API surface types from a codebase - REST routes, event handlers, CLI commands, library exports. Classifies stability and generates a complete reference. No Swagger decorations needed. Read-only. Audience: Both. Trigger: /api-survey
Architecture visualizer: maps module dependencies, detects layer violations, circular dependencies, entry points, and computes structural health score. Generates Mermaid diagrams. Read-only. Audience: Both. Trigger: /arch-vis
Changelog narrator: reads the diff between two tags/commits, clusters changes into logical groups (features, bugfixes, refactoring, dependencies), and writes a semantic changelog with breaking changes, migration notes, and deployment risk. Collector extracts git diff and clusters by module; LLM classifies each change as feature/bugfix/chore/refactor/breaking, writes migration notes, and assesses deployment risk. Read-only. Audience: Both. Trigger: /changelog
Extracts the full configuration surface from code (env vars, config files, CLI flags) and generates a human-readable reference with descriptions, types, defaults, and effects. Read-only. Audience: Both. Trigger: /config-docs
Extracts implicit coding conventions from code patterns: naming style, import style, async patterns, error handling, null handling. Quantifies each convention by consistency score and surfaces deviations. Read-only. Audience: Both. Trigger: /conventions
Traces data flow from input sources (API endpoints, events, files) through transformations to sinks (DB, APIs, logs, filesystem). Generates Mermaid flow diagrams per data flow with origin, schema changes, validation gaps, and security relevance. Read-only. Audience: Senior. Trigger: /dataflow
Domain narrator: reads all public symbols in a codebase, clusters them by call-graph density into business domains, and writes plain-English descriptions of what each domain does. Collector extracts public symbols and call graphs; LLM per cluster produces domain name, business responsibility (1-2 sentences), and business rules extracted from code. Read-only. Audience: Both. Trigger: /explain
Strategic overview of how a project handles errors: catch-type taxonomy (log/rethrow/swallow/recover/fallback), global handlers, error class hierarchy, clustered weaknesses. Read-only. Audience: Both. Trigger: /errors-overview
Maps every external integration from code: HTTP APIs, databases, message queues, storage. Classifies protocol, auth type, criticality, retry/fallback coverage. Read-only. Audience: Senior. Trigger: /integrations
Knowledge testament: mines git blame/log to map where one developer's exclusive knowledge lives (sole-author hotspots, high-churn areas they own), generates a targeted interview asking exactly the questions nobody would know to ask, and writes a structured, code-linked testament document. Read-only towards the repo. Trigger: /testament
Analyzes the topology of a codebase and generates a didactically sequenced reading tour with comprehension questions and first safe tasks. Trigger: /onboarding-pathfinder
Runbook automator: generates a deployment runbook from docker-compose.yml, package.json scripts, CI/CD config, healthcheck endpoints, Dockerfile, and README shell commands. Collector scans all five surfaces; LLM assembles a structured runbook with setup, dev workflow, deployment, healthcheck, crash recovery, CI/CD description, and troubleshooting. Read-only. Audience: Both. Trigger: /runbook
Tech-debt narrator: finds suppress comments, TODOs, empty catches, workarounds, legacy imports, and type-loosening patterns. Clusters them into logical groups and narrates repayment strategies with effort estimates. Collector scans for 6 debt types with file-level git age; LLM clusters, narrates, and estimates. Read-only. Audience: Senior. Trigger: /tech-debt
Test strategy analyzer: classifies tests into Unit/Integration/E2E, builds the test pyramid profile (ideal 60/30/10), identifies untested modules, detects overly expensive tests, and recommends optimization priority. Read-only. Audience: Senior. Trigger: /test-strategy