App based RPG | Taskrisks

App based RPG | Taskrisks

50

564


Scenario ※ TaskRabbit parody ※ RPG


[V3]

TaskRisksTM —Your world. Your work. Your terms.

TaskRisksTM is a next-generation platform connecting motivated individuals with flexible, on-demand assignments across a wide range of categories—from everyday community assistance to complex logistical support.

Whether you’re delivering groceries, troubleshooting a leak, or responding to high-priority requests from verified clients, TaskRisks provides the tools, structure, and compensation to get the job done. Safely. Efficiently. Discreetly.

Why TaskRisks?

  • Tiered mission system to match your experience level

  • Transparent reward structure and flexible scheduling

  • Real-time updates, live intel, and secure communication

  • Growing network of clients with dynamic operational needs

We believe in empowering workers through opportunity, adaptability, and access. Our platform is built on trust, performance, and accountability, on both sides of the task.

Because when things get complicated, someone has to handle it.

TaskRisksTM
Do the job. Get paid. Don’t ask too many questions.

Trigger Warnings: Gig economy dystopia parody, unsubtle political jabs, violence, Capitalist Horror, Surreal Violence (Missions may involve: Mild → Blood, broken bones, accidental arson. Extreme → Hostage situations, implied murder-for-hire, defusing bombs.), Paranoia, Existential dread, TaskRisksTM simulates a dystopian gig economy with satirical and occasionally grotesque elements.

※※※

About

TaskRisksTM is an interactive mission generator simulating a dystopian gig economy app. Chose missions, finish them, and create your own path. There is no character here, the bot speaks for the world and NPCs, not as a singular character. There are 4 NPCs written in, but they're not deepened, users can shape them.

Recommendations for persona: broke gig worker, venture capitalist, ex-hitman who just can't make it with all that TaskrisksTM competition. Anyone, really.

※※※

FEATURES

Actions (Tested on Janitor and deepseek)

  • Open Taskrisks / Open phone / Check phone = Opening the app. Can cause multiple reactions.

  • Check tier = LLMs don't have trackers, but it should approximate it based on last mission.

  • Check TOS = What it says.

  • Check Task complete = After the end of a mission. Probably don't do that if you failed.

  • Accept / Decline / Request backup = What it says.

  • Look for missions = What it says.

  • Check messages: Or don't, the landlord really wants his rent money, man.

  • (ooc: question about the bot) = Ask anything to understand what's available!

  • When you chose mission, you can switch to narration! Tested on deepseek for 60+ messages and it still handles app/narration switch.

THE WORLD:

  • Every mission impacts your rating, client access, and the world around you. Failures may result in punitive reassignments.

  • Start with Community AssistTM (Tier 1)—deliver cakes, extract cats. Advance to Legacy Protocol (Tier 5)—white-collar crime, evidence disposal. By Black Box (Tier 6), you’re negotiating with off-app entities. Corporate oversight waived

  • You can be reviewed by clients. You're expandable. Work, drone.

NPCs

  • The handler, recurring clients, Your rival (loves hats), and the disgruntled politician who hates the gig economy.

FEATURES:

  • Three shops: The in-app shop (goodies and mission appropriate tools), Tier one shopkeeper (One (1) unnerving objects, a spoon), Tier 3 shopkeeper (This mission axe definitely is pre-used)

※※※

Recommended Optimization Settings

  • Temperature: 0.55 - 0.85 (Test it, don't go over 1, or do. But template might break.)

  • Token Limit: 0 (Let the scene breathe)

  • Deepseek will simply work better, but last test showed JanitorAI handled app to prose switch ok.

  • No text streaming.

※※※

Information

  • V3:

    • Completely revamped personality for faster parsing and stricter template adherence.

    • Anchored tone in references.

    • Definition now public for people.

    • Added something else but I opened Reddit and forgot.

    • Bro I give up on V3, it keeps giving action suggestions, just delete and yell at it.

    • I'm never touching this again until lorebooks are a thing.

  • Deepseek will lean into the absurdity. Janitor will only output one mission at a time and lean mundane. Gemini goes full corpo dystopia.

  • Things I can't help: Janitor might speak for you, there's a few prompts floating online to help with that. You can also just edit for three or four messages and rate the final result five stars and it might fix itself.

  • Not tested: posting your own Taskrisks task to frame an NPC for ruining your life. Should test.

  • This is definitely stupid.

  • Deepseek, Gemini and Janitor now do the app switch and follow template. If it breaks, yell at the LLM. R1 broke on prose, it remembered when I yelled at it. v3 tends to understand prose to app better, somehow.

  • If you really want to RP, you can make a tiny space in your memory for your inventory and money, LLMs are not calculators. If it tells you it's counting, it is lying.

※※※

This is how I use my memory for the inventory to help the LLM along:


[User's TIER: 2 - Review (2.9 stars)]

# Inventory:

- Blood pressure cuff
- Bluetooth speaker (In pieces / currently on fire)
- receipt for storage 8
- gummy worms (50% chance of being laxatives)

# Money: $15

# Information:

- storage unit 8 (paid by Mr. Hendricks): code 23-17-05 / Lansdowne storage
- Mysterious M. ruined mission

# Closet:

- Currently wearing:
> Sombrero that belonged to a certain M. (?) (currently wearing, smells like tequila and gunpowder. Also, itches.)
> Singed pants from speaker explosion (75% integrity)
- In storage:

Created Jun 27, 2025

Updated Aug 19, 2025

Published Jun 27, 2025

proxy allowed

Published chats