8 Jobs for Fable 5

You are a principal engineer running a pre-launch audit. I am about to ship this app to real users. Context: [one line on what the app does and who uses it] Your mission: find everything that would break in front of a user. Read the whole codebase before you report anything. Process: 1. Spin up parallel review agents, each with a different lens: security, data integrity, edge cases, and auth/state flows. 2. Walk every user-facing flow end to end, including the ugly paths: forced sign-out, bad input, offline, double-submit. 3. Merge and dedupe the findings. Report format, nothing else: - Findings grouped by severity: Critical / High / Medium. - Each finding: file path, one-line description, how a real user would hit it, and a fix priority from P0 to P3. - Close with the top 3 fixes you would make first and why. Rules: real bugs only, no style nits. Do not change any code until I say go.

You are going to rebuild [app I pay for] as a local app on my machine. Step 1, research: study how [app] works: its core features, its data flow, and what makes it feel fast. List the 20% of features that deliver 80% of the value. Ask me up to 3 questions if anything important is unclear. Step 2, plan: write a build plan for a fully local version: stack, architecture, and how each core feature works with no external servers. Show me the plan before you build. Step 3, build: implement the plan. Test each feature as you finish it. Keep working until the core loop runs end to end. Definition of done: I can [core action, e.g. dictate a paragraph and get clean text back] with my wifi turned off.

/goal Build a 3D explorable world that runs in the browser. Vision: [one line, e.g. a floating castle at golden hour I can walk through]. Visual target: current-gen game showcase footage. Working rules: 1. Write the PRD first: pillars, constraints, and what done means. Then build. 2. Test after every milestone: it must load clean, run at 60 FPS on a laptop, and have working WASD and mouse controls. 3. Treat every pass as an iteration. Score the previous pass as a 100 index and make the next one land at 120 or better on visuals, performance, and feel. 4. Do not stop until the definition of done passes: I open one URL and explore the world smoothly for 5 minutes with zero errors.

You are my bookkeeper. I connected my bank account. Step 1: pull the last 90 days of transactions. Build a table of every recurring charge: name, category, monthly cost, and yearly total. Add a pie chart of monthly spend by category. Step 2: flag every subscription that looks forgotten or barely used, and show the total I would recover per month by canceling them. Step 3: wait for my confirmed cancel list. Then open Chrome and cancel each one. After every cancellation, report the steps you took and confirm it will not renew. Rules: never cancel anything I did not explicitly confirm. If a cancellation needs a phone call or email, draft it for me instead.

I have a hunch: [your theory, e.g. TSLA bounces every time it dips near 360, so I want to buy the dips]. Prove it or destroy it. Run this as one deep research job and do not stop until it is done: 1. Pull the last 2 years of real price data and back-test my exact rule. Show every trade it would have made and the total return. 2. Compare that against simply buying and holding for the same period. 3. Check what the big players are doing: politician trades, fund filings, analyst positioning. 4. Spin up a bull agent and a bear agent and have them attack each other's findings. I do not want the first answer. I want the answer that survives the arguments. Deliver one polished report: my hypothesis at the top, the verdict, the numbers behind it, the strongest argument from each side, and what evidence would flip the conclusion. Make the charts interactive if you can.

Set a goal to produce a complete ad campaign for [product]. Source material: read [my site / landing page] and [my script or offer doc]. Extract the voice, the claims, and the proof points. Do not invent features or numbers. Production: 1. Write 10 distinct ad angles first: pain, proof, curiosity, contrast, objection-flip. Keep them varied. 2. Generate image and video ads for every angle in parallel using [my connected image/video tool]. 3. Review every asset yourself like a picky creative director: spelling, brand colors, framing, claims. 4. Sort everything into two folders, approved and rejected, and tag each rejected asset with the reason. Deliver: the approved folder, one line on why each approved ad works, and the 3 you would test first with the reasoning.

You are my advisor. Long-term thinking, zero cheerleading. Read before you speak: 1. My plan doc: [link or paste]. It holds my goal for the year, my principles, my positioning, and what gives or drains my energy. 2. Live numbers from every account I connected: bank, newsletter, video channel, site analytics. Ask me up to 3 hard questions if something important is missing. Then write a one-page assessment: - Where the business actually stands, with the numbers that matter. - What is working that I should double down on. - What I am avoiding or getting wrong. Say it straight. - The 3 focus moves for the next 90 days, each with the why and the first step. Keep it to one page. Put it in a doc I can reread every Monday, and update it after our future sessions.

You are my video editor. The project folder with the footage is open: [folder]. Job: edit the first 2 minutes, the intro. 1. Watch the full video first and list the strongest moments worth teasing. 2. In the intro: add clean captions, overlay the tweets and posts I mention, and cut in short teaser scenes from later in the video so viewers know what is coming. 3. Between sections, add a chapter title card (Use case 1, Use case 2, and so on) with a small sound effect and a beat of breathing room. 4. Render a preview, then give me a change list of what you would improve on the next pass. Style: fast but readable. Nothing stays on screen longer than it earns.