AI coding agents are only as good as the instructions you give them. Without clear runbooks, agents guess implementation details, ignore edge cases, and invent unverified code paths.
ai-assets is an open-source repository of 23 modular AI skills, principles, and runbooks. Each skill is packaged as a token-optimized SKILL.md file that teaches your AI agent how to handle real-world engineering tasks — from Spring Boot startup debugging and Hibernate N+1 query optimization, to principal PR reviews and rapid prototyping.
Below is the complete catalog of skills, when to use them, and how to import any individual skill directly into your project using Axon.
Production-Ready AI Skills & Runbooks
Stop writing unstructured prompts from scratch. ai-assets provides 23 modular, battle-tested AI skills and constitutions for Cursor, Claude Code, Antigravity, Devin, and Codex.
Axon CLI manages and symlinks these skills across editors. Learn how Principles vs. Skills vs. Workflows work under the hood in our architectural guide.
What is an AI Skill?
An AI Skill is a structured SKILL.md spec file that combines YAML frontmatter (trigger conditions) with deterministic engineering runbooks. When loaded by your agent, it guarantees standard execution without hallucinated steps.
Metadata evaluated by the agent engine to determine when to trigger the skill based on prompt intent.
--- name: clarify-first description: Asks structured MCQs... disable-model-invocation: true ---
Deterministic instructions that guide the agent through requirements, verification checks, and safety rules.
1. Identify ambiguity 2. Present options [A/B/C] 3. Require explicit approval 4. Verify execution results
Enforces empirical verification and tags claims as Evidenced vs Assumption before finalizing response.
### 🛠️ Invoked: - clarify-first - evidence-ledger [Evidenced]
Instead of downloading skills one by one, you can clone the entire repository and bulk stage all 23 skills into your Axon hub in 2 commands:
git clone https://github.com/aayushlalroy/ai-assets.git📋cd ai-assets && axon import . --config axon-import.yaml && axon enable --all📋1. Core Engineering Primitives
Essential runbooks for clarifying ambiguity, tracking evidence, and emergency git recoveries.
clarify-first
Forces the AI agent to ask structured multiple-choice questions before executing non-trivial tasks, avoiding dangerous assumptions.
When a prompt is vague, has multiple interpretations, or involves architectural forks.
Prevents AI agents from making wrong assumptions and spending minutes generating code you didn't ask for.
The agent pauses, presents a structured list of options (A/B/C) with recommendations, and waits for your confirmation.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/clarify-first/SKILL.md -o clarify-first.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/clarify-first --name=clarify-first && axon enable clarify-first📋evidence-ledger
Forces the agent to separate verified facts from assumptions, maintaining an explicit confidence ledger.
During complex debugging sessions, root cause analysis, or performance tuning.
Stops AI agents from stating unverified hypotheses as facts.
Every claim in the response is tagged as [Evidenced: log line] or [Assumption 70%: rationale].
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/evidence-ledger/SKILL.md -o evidence-ledger.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/evidence-ledger --name=evidence-ledger && axon enable evidence-ledger📋git-recovery-runbook
Emergency safety runbook for reflog recoveries, lost commits, messy rebase fixes, and hard resets.
When you accidentally lost commits, botched a rebase, or need emergency git recovery.
Prevents accidental data loss by forcing safety backups before running destructive git commands.
The agent inspects git reflog, creates a temporary recovery branch, and restores state safely.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/git-recovery-runbook/SKILL.md -o git-recovery.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/git-recovery-runbook --name=git-recovery-runbook && axon enable git-recovery-runbook📋2. Backend & Architecture Skills
Specialized runbooks for Spring Boot debugging, Hibernate optimization, and API contract design.
spring-startup-doctor
Diagnoses Spring Boot application startup crashes, circular bean dependencies, missing configuration properties, and autowiring errors.
When a Spring Boot app fails to start with UnsatisfiedDependencyException or BeanCreationException.
Replaces trial-and-error guessing with structured stack trace inspection and bean dependency graphs.
Locates exact failing bean definition, checks profile properties, and generates targeted fix.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/spring-startup-doctor/SKILL.md -o spring-doctor.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/spring-startup-doctor --name=spring-startup-doctor && axon enable spring-startup-doctor📋hibernate-nplus1-optimizer
Identifies N+1 SELECT query bottlenecks in JPA/Hibernate entities and refactors them into set-based JOIN FETCH queries or EntityGraphs.
When database query logs reveal hundreds of repeated SELECT statements during collection rendering.
Improves database response time by 10x-100x by collapsing N+1 queries into a single query.
Analyzes entity mappings, generates `@EntityGraph` or DTO projection, and verifies query count.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/hibernate-nplus1-optimizer/SKILL.md -o hibernate-nplus1.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/hibernate-nplus1-optimizer --name=hibernate-nplus1-optimizer && axon enable hibernate-nplus1-optimizer📋openapi-contract-first
Designs REST APIs contract-first in `api-spec/` using modular `$ref` JSON schemas before writing server endpoints.
Before building new REST APIs or modifying existing backend endpoints.
Ensures frontend and backend teams align on API schemas before code implementation starts.
Generates valid OpenAPI 3.1 YAML contracts with reusable schemas and error responses.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/openapi-contract-first/SKILL.md -o openapi-contract.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/openapi-contract-first --name=openapi-contract-first && axon enable openapi-contract-first📋3. Quality & Review Orchestrators
Principal-engineer code reviews, multi-agent critique loops, and tech story drafting.
pr-review-principal
Executes rigorous, principal-engineer level code reviews analyzing API contract breaking changes, error handling, security flaws, and performance risks.
Before submitting or merging pull requests to catch subtle bugs and architectural debt.
Catches hidden regressions that generic linter tools miss (e.g. unindexed queries, breaking API schemas).
Generates a structured review report highlighting blocker issues, performance risks, and test coverage gaps.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/pr-review-principal/SKILL.md -o pr-review.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/pr-review-principal --name=pr-review-principal && axon enable pr-review-principal📋adversarial-review-loop
Multi-round Worker / Reviewer / Manager critique loop that stress-tests design proposals and implementations until 0 critical flaws remain.
For high-stakes security features, payment processing logic, or core architecture changes.
Prevents confirmation bias by pitting subagents against each other to find hidden vulnerabilities.
Subagents attempt to break the solution in multiple critique rounds until consensus is reached.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/adversarial-review-loop/SKILL.md -o adversarial-review.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/adversarial-review-loop --name=adversarial-review-loop && axon enable adversarial-review-loop📋4. Idea Lab Incubator Suite
Transform raw product ideas into validated specifications and interactive web prototypes.
Interactive MCQs
clarify-firstSteel-man & Red-team
scrutinize-ideaAdversarial Review
adversarial-review-loopReact + Vite Web App
visualize-idea-websiteidea-lab
Master 4-phase incubator workflow that guides an idea from vague prompt to clarified spec, red-teamed architecture, and interactive web prototype.
When starting a new feature, side project, or product concept from scratch.
Combines 5 primitive skills into an end-to-end pipeline so you don't skip validation steps.
Runs Clarify → Scrutinize → Harden → Generates React+Vite prototype.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/idea-lab/SKILL.md -o idea-lab.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/idea-lab --name=idea-lab && axon enable idea-lab📋visualize-idea-website
Generates production-grade, interactive React + Vite prototype web applications featuring rich aesthetics, glassmorphism, and live state logic.
When you want to demonstrate a feature idea with a live, interactive UI mockup.
Turns abstract text ideas into tangible visual web apps you can interact with immediately.
Builds a full React frontend project with components, styles, and interactive state.
curl -sSL https://raw.githubusercontent.com/aayushlalroy/ai-assets/main/skills/visualize-idea-website/SKILL.md -o visualize-idea.md📋axon add skill https://github.com/aayushlalroy/ai-assets/tree/main/skills/visualize-idea-website --name=visualize-idea-website && axon enable visualize-idea-website📋5. Always-On Principles (Constitutions)
Machine-enforced coding rules that shape agent responses across every turn.
Mandates that the AI agent ends every response with a footer listing all skills and rules invoked during the turn.
axon add principle principles/skill-attribution.md --name=skill-attribution && axon enable principle skill-attributionEnforces inline tagging of statements as [Evidenced: source] or [Assumption NN%: reasoning] to eliminate unverified claims.
axon add principle principles/claim-tagging.md --name=claim-tagging && axon enable principle claim-taggingAppends turn-by-turn token consumption breakdowns (User, Agent, Output, Total) to response footers.
axon add principle principles/token-attribution.md --name=token-attribution && axon enable principle token-attributionRequires updating docs (README, CHANGELOG, CRQ) and bumping SemVer version for all code changes.
axon add principle principles/doc-version-sync-policy.md --name=doc-version-sync-policy && axon enable principle doc-version-sync-policyExplore the Complete ai-assets Repository
All skills and principles are open-source and free to use in your engineering workflow.