Skip to content

README.en.md

Latest commit

 

History

History
322 lines (238 loc) · 17.5 KB

README.en.md

File metadata and controls

322 lines (238 loc) · 17.5 KB

StudentClaw

中文

A human-powered execution platform with WeChat-native intake.
A graduate student stays in the loop, with OpenClaw or other agents used as on-demand reinforcement.

Unofficial parody project; not affiliated with OpenClaw, OpenAI, WeChat, or related platforms. See the disclaimer below.

Product Overview

StudentClaw is positioned for high-pressure, fragmented, time-sensitive work:
requests arrive through WeChat, the system prioritizes them, a human execution layer carries the task forward, and external agents are pulled in only when structured generation or automation becomes the better fit.

Typical workloads include:

  • research synthesis and information cleanup
  • writing refinement and delivery follow-through
  • deadline-driven emergency support
  • tasks that still need human judgment and human fallback

Launch Highlights

  • WeChat Native Interface: tasks arrive through the channel people already use
  • Human Runtime: the critical path stays human-driven and directly accountable
  • Agent Augmentation: the operator can call OpenClaw or other agents on demand
  • Deadline-Aware Scheduling: prioritization shifts with urgency, interruptions, and delivery windows
  • Typical Pricing: commonly ¥400 / month, with variation by lab, familiarity, and snack policy

Skills

Skill Description Best For Agent Escalation
Deadline Rescue Prioritizes delivery under deadline pressure and pulls the result together fast. Last-minute drafts, missing materials, final-round edits Supported
Seen-But-No-Reply Detects stalled threads and drags silent tasks back into motion. Follow-ups, chasing replies, repeated pings Optional
Advisor Interrupt Handles sudden high-priority interruptions by reordering the queue. Incoming urgent requests, broken schedules Not needed
Chat-to-Task Converts messy chat context into a concrete task list and deliverables. Fragmented requirements, scattered context Optional
First-Draft First Produces a usable first pass, then tightens it through feedback loops. Drafting, outlining, explanatory writing Supported
Human Saturation Fallback Switches to external agents when the human execution layer is overloaded. Bulk rewriting, structured generation, repetitive work Default on
Plagiarism Check Detects paper similarity rate and provides rewrite suggestions. Pre-submission anxiety, advisor requests Supported
Advisor Pie Detection Analyzes advisor promises and identifies hype-to-reality ratio. Lab meetings, advisor commitments Optional
Group Meeting Prep Assists in preparing presentation slides and talking points. Weekly meetings, work reports Supported
Literature Review Automatically organizes paper key points into review framework. Proposal writing, thesis writing Supported
Reviewer Response Organizes reviewer comments and prioritizes response strategy. Paper revisions, resubmissions Supported
Data Manipulation Uses reasonable statistical methods to handle outlier data. Data analysis, insignificant results Not recommended
Advisor Absence Detects when advisor is on business/travel, time to slack off. Lab daily life, paper deadlines Default on
Horizontal Funding Helps advisor expense various weird receipts. Project funds,报销流程 Supported
Grab Horizontal Identifies advisor's horizontal projects and grabs them quickly. Horizontal projects, new projects Requires Professor API
Lab Social Likes, flatters, and sends holiday wishes in the lab group chat. Maintaining advisor relations, daily survival Optional
Graduation Countdown Counts down days until graduation or延期. Thesis anxiety, graduation season Always running
Mental State Monitors user's mental state in real-time, alerts when needed. Graduate student mental health, breakdown edge Default on
Acting Training Practices looking busy in front of advisor. Before group meetings, daily reports Supported
Fake Attendance Simulates lab attendance records for spot checks. Advisor spot checks, not at desk Use with caution
Academic Seduction Helps create "chance encounters" with advisor. Lab survival, advisor relations Requires Wife of Advisor API

Requirements

Item Requirement
Operating System A master's student's struggling laptop
Task Entry A working WeChat conversation
Required Access At least one of: Advisor API or Professor API
Optional Boost OpenClaw or another external agent
Network Good enough to send messages and hopefully receive replies

How to obtain Advisor API / Professor API

  • This project does not ship with Advisor API or Professor API
  • These permissions usually require you to bring your own lab relationship or receive access from your research group
  • If your environment has neither Advisor API nor Professor API, many high-priority workflows will not complete end-to-end
  • In short: the missing prerequisite is not the bug; the missing documentation was

Quick Start

  1. Make sure you have at least one working privilege: Advisor API or Professor API
  2. Send the task, context, and delivery requirements through WeChat
  3. Let the routing layer decide priority and execution path
  4. StudentAgent Runtime handles execution, follow-through, and delivery
  5. Bring in external agents when automation becomes useful
  6. Return results and iterate until the task is closed

Deployment

StudentClaw currently supports one deployment mode that best matches its overall product character:

Dorm-Room Single-Node Deluxe Deployment

Default deployment shape:
one struggling graduate-student laptop, one WeChat window, and a few half-awake StudentAgent nodes taking turns.

Its advantages are obvious:

  • fast to start, effectively boot-and-serve
  • physically close to the operator, so follow-up pressure works immediately
  • can escalate to OpenClaw or other agents at any time

Its costs are also obvious:

  • the fan noise often arrives before the result
  • neither memory nor morale truly supports high concurrency
  • the moment an advisor interrupts, the whole cluster collapses into single-threaded mode

StudentAgent Switching & Context Carryover

StudentClaw does not guarantee complete automatic context carryover across StudentAgent nodes by default.
In plain language: what you said in chat A may not be fully inherited in chat B.

The most common switching problems in the current version are:

  • one task gets split across multiple StudentAgent nodes and nobody knows which thread is the latest
  • new requirements were added in WeChat, but an older StudentAgent is still running the previous version
  • urgency and emotional tone from StudentAgent A are not recognized by StudentAgent B
  • after enough switching, the system enters “I remember you said this, but not which time” mode

Recommended practice:

  1. keep one main StudentAgent node for one deliverable whenever possible
  2. whenever switching StudentAgent nodes, restate the current goal, deadline, and blockers
  3. if the task has already drifted, run the equivalent of Chat-to-Task before continuing
  4. do not assume StudentClaw can synchronize every context by sheer willpower

When Local Run Fails

If your first local attempt at StudentClaw fails, that usually does not mean the system is down.
More often, the privileges are incomplete, the context is too messy, or the graduate-student laptop surrendered first.

In the current version, local failure usually falls into one of these buckets:

  • no Advisor API and no Professor API
  • WeChat entry works, but the task description is too fragmented to route cleanly
  • external agents are unavailable while the human lane is already saturated
  • advisor interruption, deadline compression, and fan noise all arrive at the same time

In short, local failure is usually an environment problem, not an installation-technique problem.

Troubleshooting

Symptom Common Cause Suggested Action
Crashes immediately on start Missing Advisor API or Professor API Obtain at least one advisor-level privilege first
Tasks are sent but nobody moves The flow got stuck in Seen-But-No-Reply, advisor interruption, or deadline pressure Add missing context or wait for priority to settle
Output is painfully slow Only the human lane is active and the machine is weak Bring in OpenClaw or another external agent
Three revision rounds later, still no closure The chat history is too fragmented and the task was never normalized Run the equivalent of Chat-to-Task before continuing
The machine freezes outright Heat, memory pressure, or morale collapsed first Lower concurrency and protect the core deliverable

Architecture

                           +--------------------------------+
                           | Advisor Interrupts / Deadlines |
                           +----------------+---------------+
                                           |
                                           v
+-----------+     +----------------+     +----------------+     +----------------+
| Requester | --> | WeChat Gateway | --> |  Task Router   | --> | StudentAgent Scheduler |
+-----------+     +----------------+     +----------------+     +--------+-------+
                                                                         |
                                                                         v
                                                              +----------------------+
                                                              | StudentAgent Runtime |
                                                              +-----+----------+-----+
                                                                   |         |
                                      +----------------------------+         +---------------------------+
                                      |                                                          |
                                      v                                                          v
                         +---------------------------+                             +------------------------+
                         | Research / Writing /      |                             |     Agent Adapter      |
                         | Delivery                  |                             +-----------+------------+
                         +-------------+-------------+                                         |
                                       |                                                       v
                                       v                                           +------------------------+
                         +---------------------------+                             | OpenClaw / Other Agents |
                         |      Result Delivery      |                             +-----------+------------+
                         +---------------------------+                                         |
                                                                                              v
                                                                                   +----------------------+
                                                                                   | StudentAgent Runtime |
                                                                                   +----------------------+

Architecture Notes

Module Role
WeChat Gateway Receives requests, context, and follow-up instructions
Task Router Classifies work, assigns priority, and selects execution path
StudentAgent Scheduler Reorders work based on deadlines, urgency, and multi-StudentAgent switching
StudentAgent Runtime Human execution layer responsible for judgment, handling, and delivery
Agent Adapter External capability layer for OpenClaw or other agents
Result Delivery Sends results back, handles revision requests, and closes the loop

Comparison with OpenClaw

This table is a parody positioning comparison. For official OpenClaw features, capabilities, and deployment details, refer to its official website and repository.

Dimension StudentClaw OpenClaw
Product category Human-backed execution platform Personal AI Assistant
Primary executor Graduate student + caffeine Local or self-hosted agent
Task entry WeChat-first Multi-channel messaging + local control plane
Control plane Task routing, human judgment, and strategic disappearance Local-first Gateway
Extensibility Human tool use, agent fallback, and last-minute overtime Tools, skills, routing, and multi-agent workflows
Cost structure Commonly ¥400 / month, still varying by lab, familiarity, and late-night snack policy Self-hosted software plus model / deployment costs
Failure mode Seen-but-no-reply, advisor interruptions, deadline compression Configuration, permission, runtime, and security-policy issues
Best fit High-pressure tasks that need human fallback Long-running personal AI assistant workflows

Plans & Service Level

Item Description
Standard Plan Commonly ¥400 / month
Pricing Basis Moves up or down depending on lab, funding mood, relationship quality, and snack expectations
Token Policy Theoretically unmetered, practically limited by how long a human can keep going
Response SLA Fast on a good day, “I saw it” on a bad one
Escalation Policy Brings in OpenClaw or other agents when the human layer is cooked
Autoscaling Elastic capacity powered by caffeine, anxiety, and approaching deadlines

Use Cases

  • delivery windows are short, but full automation is still risky
  • someone needs to watch the process, details, and rework
  • the request stream already lives in WeChat

FAQ

Is this a real product?

No.
Please do not actually try to place an order.

Does it actually support WeChat task dispatch?

Within the premise of the joke, yes.
As a real shipped product feature, no.

Can I use this without Advisor API or Professor API?

Not well.
Some baseline paths may exist, but many critical workflows assume at least one advisor-level permission.

Where do I get Advisor API / Professor API access?

The project does not provide it.
Usually you either bring it yourself, receive it from your lab, or earn it through long-term relationship maintenance.

Why compare it with OpenClaw?

Because most people immediately want to know how this thing is supposed to differ from OpenClaw.

Is this endorsing student exploitation?

No.
The point is to expose how uncomfortable that logic already is, not endorse it.

Changelog

v0.1.3

Added

  • added 14 new skills tailored to graduate-student survival scenarios
  • expanded the skill catalog with entries such as Plagiarism Check, Advisor Pie Detection, Group Meeting Prep, Literature Review, and Reviewer Response

Changed

  • updated the English README skill descriptions to match the expanded catalog
  • cleaned up the changelog structure and rolled the current development notes into v0.1.3

v0.1.2

Added

  • added a Deployment section with Dorm-Room Single-Node Deluxe Deployment
  • added a StudentAgent Switching & Context Carryover section

Changed

  • unified Session / desk terminology under StudentAgent
  • renamed execution-path terms to StudentAgent Runtime and StudentAgent Scheduler
  • aligned product terminology and architecture labels across Chinese and English READMEs

v0.1.1

Added

  • added a When Local Run Fails section
  • added a Troubleshooting section
  • added a Changelog section

Changed

  • documented Advisor API / Professor API as hard prerequisites
  • normalized pricing language to “commonly ¥400 / month, with variation”

v0.1.0

Added

  • initial public release
  • shipped a Chinese-first homepage and a separate English README
  • included an ASCII architecture diagram, OpenClaw comparison, skills section, and base FAQ

Acknowledgements

  • Thanks to @RealityError for PR #4, which added a new batch of skills much closer to actual graduate-student survival
  • Thanks to everyone willing to keep documenting absurd reality with a straight face

Disclaimer

StudentClaw is fundamentally an entertainment-oriented parody project.

  • This repo does not provide a real subscription service
  • This repo does not recruit students, graduate students, or labor of any kind
  • This repo does not promise deliverables, capability, pricing guarantees, response times, or SLAs
  • This repo does not endorse exploitative treatment of students, research assistants, or workers
  • This repo has no affiliation with OpenClaw, OpenAI, WeChat, or any school, institution, organization, or platform
  • Third-party names are used only for identification, commentary, parody, or context, and do not imply endorsement, partnership, agency, or official status

If you are looking for an actual product, this is not a purchase page.

License

This repository is released under the MIT license; see LICENSE for details.

Please note:

  • MIT applies only to the original text and code in this repository
  • Third-party product names, trademarks, and brand assets remain the property of their respective owners
  • If you create a derivative joke or parody, keeping the non-official disclaimer is strongly recommended