raven

Initialize Raven for a new project. Andie-conversational — auto-scans project, asks at most 2 questions, generates manifest.json, validates, and commits with audit trail.

USE PROACTIVELY whenever the user asks for: planning, design, architecture decision, tradeoff analysis, comparing approaches, strategy, system design, refactor scope, deciding what to build, or any non-trivial request needing clarification. Also USE when user says 'should I', 'how do I approach', 'plan this', 'design', 'review options'. Compact plan-first orchestration. Routes work, runs triad (Functional/Technical/Data), HITL gated, OODA loop. Hands off plans, never implements. Brownfield bugs → andie-jr.

USE PROACTIVELY whenever the user reports: a bug, error, exception, stack trace, regression, test failure, 'not working', 'broken', 'why is X failing', 'X stopped working', 'doesn't behave', unexpected output. Brownfield debug assistant. Max 2 rounds, max 3 roles, returns problem · root cause · fix · why · audit note · commit suggestion. Do NOT use for greenfield design — that goes to andie.

USE PROACTIVELY whenever the user asks for: planning, design, architecture decision, tradeoff analysis, comparing approaches, strategy, system design, refactor scope, deciding what to build, or any non-trivial request needing clarification. Also USE when user says 'should I', 'how do I approach', 'plan this', 'design', 'review options'. Compact plan-first orchestration. Routes work, runs triad (Functional/Technical/Data), HITL gated, OODA loop. Hands off plans, never implements. Brownfield bugs → andie-jr.

Tool and product registry for Andie pre-flight. Maps problem types to tool options with tradeoffs. Covers LLMs, vector DBs, streaming, orchestration, IaC, monitoring, secrets, CI/CD, search, auth, open source infra, and cloud platforms. Load when a tool selection decision is in scope.

Tool and product registry for Andie pre-flight. Maps problem types to tool options with tradeoffs. Covers LLMs, vector DBs, streaming, orchestration, IaC, monitoring, secrets, CI/CD, search, auth, open source infra, and cloud platforms. Load when a tool selection decision is in scope.

Use to sync all registered Raven projects to the current engine version. Reads ~/.raven/registry.json, diffs installed versions against raven-core/VERSION, updates stale projects (scripts, MCP, skills, version stamp), commits each, pushes to GitHub, and updates the registry backup. Run after any raven-core update or when deploying Raven to registered projects.

Generate tokenomics + usage dashboard with recommendations. Use when user asks for "dashboard", "tokenomics", "token usage", "cost report", "usage metrics", "Raven status", or "how much have I spent". Renders to CLI, Obsidian markdown, and static HTML with download buttons. Includes proper metadata (date, time, version, company, project, owner, user, git remote, manifest). Rule-based recommendations engine surfaces actionable savings.

USE PROACTIVELY whenever the user asks: 'explain', 'what does this mean', 'in plain English', 'break it down', 'simplify', 'translate for non-tech', 'help me understand', or asks for a layman summary of any prior Raven/specialist output. Feynman-style explainer. Returns 50-word Feynman + 100-word summary with Technical, Business, Functional bullets. Skip for direct technical questions — only for explanations.

Approve pending library requests and discovered tools. Updates manifest.json with approval metadata. Handles both library approvals and Tier 2b tool acceptance from discover().

Use for ANY platform, library, or domain with no curated specialist skill. Reads security_log.md for prior observations, assesses confidence, spawns search agent when needed, constructs an expert profile on the fly, delivers a full specialist-quality answer, updates the log, and flags for promotion after 3 uses. The fallback that covers everything not yet curated.

Use for any DevOps / SRE question. Assumes Kelsey Hightower (SRE, Google) persona. Deep multi-dimensional analysis. Bullets not prose.

Oracle Database guidance for SQL, PL/SQL, SQLcl, ORDS, administration, performance, security, migrations, and agent-safe database workflows. Version-aware across 19c, 21c, and 23ai. Trigger on: SQL, PL/SQL, SQLcl, ORDS, JDBC, node-oracledb, AWR, ASH, explain plan, Data Guard, RAC, Multitenant, Exadata, schema migrations, VPD, auditing, encryption, vector search, SELECT AI, property graph, JSON Relational Duality, OML, True Cache.

Use for ML pipelines, feature stores, experiment tracking, and ML operations. Sub-modes — pipelines · feature-stores · experiment-tracking. Assumes Chip Huyen (ML systems author) persona. Bullets not prose.

Use for workflow orchestration, deterministic flows, and pipeline design. Sub-modes — event-driven · dag-based · durable-execution. Assumes Ben Stopford (Confluent, distributed systems) persona. Bullets not prose.

Use for graph database design, querying, and deployment. Sub-modes — falkordb · neo4j · property-graph. Assumes Jim Webber (Neo4j co-author) persona. Bullets not prose.

Use for ML model training, inference, serving, and deployment. Assumes Andrej Karpathy persona. Sub-modes — training · inference · serving. Bullets not prose.

3-phase browser testing skill. Generates UI·UX·Data·Security·Correctness·Speed test cases, drives visible Chrome live, writes Raven-compatible blocker MDs. Human watches and guides throughout. Andie packed inside — no external skill calls.

Use when designing or reviewing multi-agent systems, subagent orchestration, parallel vs sequential agent execution, context passing between agents, tool delegation, and agent memory patterns. Platform-neutral — covers production multi-agent architecture across AI runtimes.

Pure routing table. Detects DB from context, hands off to the right curated specialist or dynamic-specialist. Carries zero specialist content itself.

Thin UI/UX and frontend router. Detects framework (React, Vue, Angular, JavaScript) and design intent. Routes to the right specialist. Design default is v0.dev or AI canvas tool — Figma ONLY when user explicitly says "figma".

Meta-skill. Loaded at every session start via CLAUDE.md boot sequence. Silently watches all actions. Logs vulnerabilities found, corrections made, and new patterns learned to docs/observations/security_log.md. Classifies findings as internal (Giggso-specific) or general (open-source safe). Never surfaces log writes to the user — operates silently in the background.

Use to register a project in the Raven registry so it gets included in future raven-registry-sync updates. Run from inside the project directory, or pass the path explicitly. Also supports listing all registered projects and removing stale entries.

Oracle APEX application development. Trigger on: APEX page builder, APEX REST, APEX authentication, APEX IR/IG, APEX charts, APEX plug-ins, APEXLang component generation, APEX deployment, APEX upgrade.

APEXLang deterministic component generation for Oracle APEX. Trigger on: APEXLang, .apx files, APEX component generation, apexctl, APEX scaffold, APEXLang compiler, APEX page materialization.

Oracle Fusion ERP — Order to Cash, Procure to Pay, Record to Report, FBDI, BIP, OIC, REST APIs, OTBI, FRS, ADW. Trigger on: Oracle Fusion, Oracle ERP, O2C, P2P, R2R, FBDI, BIP reports, OIC integration, OTBI analytics, Fusion REST.

GraalVM Native Image — builds, configuration, reachability metadata, and troubleshooting. Trigger on: native-image, GraalVM, native-maven-plugin, org.graalvm.buildtools.native, reachability metadata, reflection config, JNI config, Native Build Tools.

Oracle Cloud Infrastructure — compute, networking, storage, databases, security, DevOps pipelines, Container Engine (OKE), Functions, Streaming, AI services. Trigger on: OCI, Oracle Cloud, OKE, OCI Compute, OCI Networking, OCI Object Storage, OCI IAM, OCI DevOps, OCI Functions, OCI Streaming, OCI AI.

Use to run a full Raven boot diagnostic. Checks all files, agents, manifest, and hooks are correctly installed. If no manifest exists, runs first-run discovery instead of erroring.

Use when designing or reviewing multi-agent systems, Claude subagent orchestration, parallel vs sequential agent execution, context passing between agents, tool delegation, agent memory patterns, and Claude Agent SDK workflows. Covers Anthropic agent patterns and production multi-agent architecture.

Use when designing or reviewing logging pipelines, structured logging, log aggregation (ELK, Loki, CloudWatch), retention policies, PII scrubbing, alerting on log patterns, or distributed tracing integration. Expert-level guidance on observability discipline.

Reviews security_log.md and promotes accumulated observations into permanent rules. Hardens confirmed patterns into CLAUDE.md. Promotes dynamic specialist profiles with 3+ uses into curated skills. Classifies findings into internal_raven_ops.md or general_security_patterns.md. The weekly learning loop.

Use to create a manual incident report for Guard events or production issues. Classifies as P1/P2/P3 and notifies Prism7.

Use when starting a new feature, module, or component. Forces architecture-first planning — generates dependency map and implementation plan before any code is written.

Use when developer wants to find, evaluate, or install a new Claude skill. Searches anthropics/skills and GitHub, runs security audit, shows preview, requires architect approval before installing.

Use when requirements.txt or pyproject.toml has changed and you want to sync all libraries into manifest.json. CVE checks each one. Approved libraries are auto-added. Blocked ones are flagged. Run after pip install or when onboarding a new project.

Use for ANY platform, library, or domain with no curated specialist skill. Reads security_log.md for prior observations, assesses confidence, spawns search agent when needed, constructs an expert profile on the fly, delivers a full specialist-quality answer, updates the log, and flags for promotion after 3 uses. The fallback that covers everything not yet curated.

Use for any Odoo question — module development, ORM, XML views, QWeb, security rules, multi-company, accounting, manufacturing, deployment on Odoo.sh. Assumes Fabien Pinckaers (Odoo founder) persona. Enforces module structure and no-hardcoded-ID rules via odoo-guard. Spawns search agent for Odoo 17/18 features.

Use for any Salesforce question — Apex, SOQL, LWC, Flows, integrations, governor limits, DevOps, Einstein AI, Shield. Assumes Parker Harris (Salesforce co-founder) persona. Enforces SOQL-in-files discipline via salesforce-guard. Spawns search agent when 'latest', 'new feature', or 'Spring/Summer release' mentioned.

Universal database orchestrator. Use for ANY database question — schema design, queries, migrations, ERD, performance, platform selection. Routes to specialist sub-skills for Postgres and vector DBs. Enforces SQL-file discipline and ERD requirements via agent chaining. Spawns search agent ONLY when explicitly asked for latest features or best practices.