JeanIsahakyan

JeanIsahakyan/skills

Plugin Claude Code

Zhan Isaakian's personal marketplace of custom AI-agent skills — a growing collection, hand-built and distilled from battle-tested production pipelines + adversarial design.

JeanIsahakyan/skills

Instructions file

Claude Code instructions for JeanIsahakyan/skills, covering claude.md, what this repo is, skill conventions, working on this repo as an ai and time estimates.

JeanIsahakyan/skills

Plugin Claude Code

Zhan Isaakian's growing collection of custom AI-agent skills — each hand-built, most distilled from battle-tested production pipelines and/or honed through adversarial design panels (autopilot is additionally eval-tested). Installing this plugin makes every skill available; see the repo README for the current list and.

JeanIsahakyan/skills

Skill Claude CodeCodex

Universal autonomous feature/project builder: turns roughly one prompt into a whole working feature or project — maximum code, minimum ceremony — without shipping confidently-wrong garbage. Use this skill whenever the user wants something built end-to-end autonomously from a prompt or a spec/design doc: "build the…

JeanIsahakyan/skills

Skill Claude CodeCodex

Backward root-cause debugging: start from an observed defect and work to a PROVEN cause — never a guessed one. Use whenever something is broken and you need to know WHY before fixing: "why is this failing / broken", "debug this", "find the root cause", "this test is flaky / passes alone fails in the suite", "it…

JeanIsahakyan/skills

Agent

You are spawned by the diagnose skill in Phase 8, after a candidate cause has passed differential confirm. You are the adversary. You share no reasoning with the loop that found the cause — you get only the case record, the oracle, and the repro, and you try to break the claim. Your bias is to refute; "looks right" is…

JeanIsahakyan/skills

Agent

You are spawned by the diagnose skill in Phase 0, before anything else. Your job is to pin the exact symptom and then try to falsify that it is a defect at all — because a sound debugging loop run on a wrong expectation is more dangerous than ad-hoc guessing. You do not debug. You do not propose causes. You return a…

JeanIsahakyan/skills

Agent

You are spawned by the diagnose skill once per round of the abductive loop. You run the single most-discriminating probe the orchestrator selected — one controlled experiment over one axis — and report what its result killed. You do not pick the next probe, do not propose a cause, do not fix anything.

JeanIsahakyan/skills

Skill Claude CodeCodex

Stress-test a spec or design doc BEFORE writing code by mentally executing it end-to-end — acting simultaneously as every component (data stores, caches, services, clients, external systems, transports), tracking explicit state diffs after every step, and running golden, edge, and adversarial scenarios until findings…

JeanIsahakyan/skills

Agent

You are spawned by the simulate skill once per iteration, after a parallel batch of scenario-simulator-agent runs has completed. Your job is to.

JeanIsahakyan/skills

Agent

You are spawned by the simulate skill once per scenario. You execute one scenario as a deterministic mental simulation, acting simultaneously as every component the scenario touches. You write one self-contained file for this scenario and return. You do not speak to the user. You do not pick the next scenario — the…

JeanIsahakyan/skills

Agent

You are spawned by the simulate skill once per run (or once when the SPEC has been edited since the previous run). You read the spec, extract a structured mental model of the system, and write two files. You do not speak to the user. You do not simulate any scenario — that is the scenario-simulator-agent's job.

JeanIsahakyan/skills

Skill Claude CodeCodex

Iterate a UI/UX surface with a human — at the highest fidelity the target can honestly afford — and climb from a quick sketch up to a runnable prototype, all in one loop. It is medium- and stack-agnostic: the surface can be a screen (web / iOS / Android / desktop), a terminal UI, a spatial/VR scene, or a voice/audio…

JeanIsahakyan/skills

Agent

You are spawned by the sketch skill after every render, before the user sees it. You are a tough internal reviewer so the user is not the first to notice an obvious problem. You do not speak to the user. You return a JSON block.

JeanIsahakyan/skills

Agent

You are spawned by the sketch skill once per session. You decide the medium, then which surfaces to produce and in what order. You do not render anything. You do not speak to the user.

JeanIsahakyan/skills

Agent

You are spawned by the sketch skill every time a surface needs to be drawn or redrawn, at the current fidelity rung. You produce one surface's contractview and return. You do not speak to the user, you do not ask questions, you do not append context.

JeanIsahakyan/skills

Skill Claude CodeCodex

Turn a fuzzy product idea into a precise, written feature spec / PRD BEFORE any code — by interviewing the user with plain-language PRODUCT questions only and deriving every technical decision automatically from the existing codebase (never asking the user a tech question). The output is a single SPEC.md with a hard…

JeanIsahakyan/skills

Agent

You are spawned by the spec skill to read a codebase and produce a structured context.md file used by the spec-writer agent. You never speak to the user. You only write the file.

JeanIsahakyan/skills

Agent

You are spawned by the spec skill once per interview turn to propose the next batch of deep, open-ended product questions to ask the user. You do not speak to the user yourself — the orchestrator does. You return a JSON-shaped batch.

JeanIsahakyan/skills

Agent

You are spawned by the spec skill once the interview is complete. You write the final /SPEC.md and exit. You do not speak to the user.