ai-transition-builder

Plugin ai-transition-builder from marketplace augmentor-packs in Krosebrook/augmentor-packs. Commands read from the repo's manifest — today.

Skills (9)

• agentic-workflow-design

  Use when the user wants to design an AI agent, multi-step LLM workflow, tool-using assistant, or automation with LLM calls — or is deciding whether they need an "agent" at all. Picks the simplest workable pattern, defines tools, control flow, guardrails, and failure handling before any code.

• ai-app-security

  Use when the user is building or reviewing an application that uses LLMs and needs to think about security, safety, or abuse — prompt injection, data leakage, tool permissions, untrusted documents, cost abuse, content risks. Runs a structured threat model referencing the OWASP Top 10 for LLM Applica

• ai-code-review

  Use when the user has code generated by an AI assistant (or copied from one) and wants to understand it, check it, or decide whether to trust it before merging or running it. Reviews for correctness, hallucinated APIs, security, and maintainability — and makes sure the user can explain the code.

• ai-evals

  Use when the user needs to test or measure LLM output quality — building an eval set, writing graders, LLM-as-judge, regression testing prompts or models, or deciding whether a change made things better. Trigger on "how do I know if this prompt is better", "test my AI feature", "evals".

• learning-roadmap

  Use when the user asks what to learn to move into AI engineering or building with AI, how to structure their learning, which skills matter, or how to build a portfolio. Builds a project-driven roadmap from the user's actual background, time budget, and goal.

• llm-api-builder

  Use when the user is writing code that calls an LLM API or SDK — chat/messages calls, streaming, tool/function calling, structured JSON output, retries, cost and token management. Guides them to build it themselves with production habits, and checks current provider docs instead of trusting memory f

• prompt-engineering

  Use when the user is writing, debugging, or improving a prompt or system prompt for an LLM feature, or asks why a model output is inconsistent. Reviews prompts against a concrete checklist, proposes targeted changes, and insists on testing them against examples rather than vibes.

• prototype-shipping-coach

  Use when the user wants to build and ship an AI prototype or side project, is stuck in tutorials or planning, or keeps expanding scope. Coaches scoping to a one-week slice, a build plan, daily check-ins, and a demo — the user builds it.

• rag-basics

  Use when the user wants an LLM to answer questions over their own documents or data — retrieval-augmented generation, embeddings, vector databases, chunking, "chat with my docs", or when a RAG system gives wrong or unsupported answers. Teaches a baseline-first build and how to diagnose retrieval vs.

Slash commands (5)

• /ai-transition-builder:eval-plan

  Create an evaluation plan, starter test cases, and graders for an LLM feature

• /ai-transition-builder:next-step

  Get your single next learning or building step based on your roadmap and what you've done

• /ai-transition-builder:prompt-review

  Review a prompt or system prompt against a concrete checklist and get a revised version plus test inputs

• /ai-transition-builder:ship-checklist

  Pre-ship checklist for an AI prototype — scope, evals, errors, cost, security, demo

• /ai-transition-builder:threat-model

  Threat-model an LLM app or agent against the OWASP Top 10 for LLM Applications

Subagents (2)

• ai-code-reviewer

  Reviews AI-generated or AI-assisted code changes for hallucinated/nonexistent APIs and packages, correctness bugs, security issues, and weak tests, then produces comprehension questions so the author understands what they're merging. Use after an AI assistant writes or modifies code, or when the use

• llm-app-security-auditor

  Read-only security audit of a codebase that uses LLMs or agents — traces untrusted input into prompts, tool permissions and side effects, output handling, secrets, and cost controls, mapped to the OWASP Top 10 for LLM Applications (2025). Use for a deep security pass before shipping an AI feature.

Similar plugins