Skip to content
Limin Zheng

AI Skills & Workflows

AI workflow & automation

Define a small sequence of AI and deterministic steps, with approvals where needed.

Choose this when
Choose this for governed workflows across systems/triggers; for standalone scripts choose Local scripts & small utilities.
Scope unit
One workflow, triggers and systems involved
Possible delivery format
Feasibility/permission report first; workflow and tests after approved implementation
Prepare a brief →

You send · You receive

What to send

  • Trigger, input, destination systems and desired outcome
  • Rules, failure paths and human approval steps
  • API access, sandbox, run frequency and cost ceiling

Start by describing the materials you already have; missing details can be added when we agree the scope. Before work begins, we will confirm the materials, permissions and working conditions needed for your specific task.

Two stages

Stage 1 · Diagnosis

First receive a workflow feasibility and permission assessment covering triggers, source/destination systems, API/sandbox availability, human approvals, failure paths, proposed implementation scope and blockers.

Possible outputs

  • Feasibility/permission report, proposed flow, dependencies/blockers and recommended scope

A diagnosis can be the complete order.

You decide whether to continue

Stage 2 · Approved implementation

Implementation is separately scoped and quoted once interfaces, limited authorization, a test environment, frequency and cost ceilings are confirmed. Scripts/configuration, executed tests, deployment or operations are delivered only within their approved scope.

Possible outputs

  • Flow diagram, scripts/configuration and exception runbook
  • Normal/failure/approval tests in the agreed environment

Scoped and quoted separately.

Third-party costsThird-party APIs, model runs, account subscriptions, hosting and ongoing operations are separate costs to agree in advance, not implicitly covered by the delivery fee.

Key checkDiagnosis: explain feasibility, blockers, permissions and next-step scope; test normal/failure/approval flows after approved implementation.

Verifiable acceptance · 3
  • Stage 1 · DiagnosisThe report lets the client assess feasible steps, access/API prerequisites, human approvals and unverified items
  • Stage 2 · Approved implementationExecute agreed normal/failure paths and report results and gaps
  • Stage 2 · Approved implementationPermission/approval tests for external writes/sends, with costs, frequency and maintenance ownership documented

Hypothetical brief & output structure

Illustration · hypothetical

Illustrative example — not a completed client order

The request

Classify selected emails into CSV, with mandatory human approval before any outbound send.

Sample artifacts

  • workflow-assessment.mdfeasibility, proposed flow and access/dependency boundaries
  • automation/only after separately approved implementation
  • tests.mdsandbox execution evidence after approved implementation

File names are illustrative

Other tasks of this kind

  • Convert newly added authorized files in one folder into rows each day; require approval for external sending.
Output structure / test path

Flow: trigger → filter → classify → human approval → save / optionally send

Write a brief from this example →

Scope boundaries

Confirm APIs, account permissions, approvals, writes and operating costs first; no ongoing custody of production secrets.

Not included by default

  • Automatic unrestricted access to production data
  • Sharing live customer passwords or tokens
  • Perpetual operation on my personal accounts

What affects the quote

Separate diagnosis, implementation, deployment and operating costs.

Actual project references

  • Roleward ↗

    Self-directed AI product

    Workflow and product practice, not evidence of market success or client delivery.

  • Asterline ↗

    Self-directed AI workflow and eval with synthetic data

    Evaluation on synthetic test data is not real-client business accuracy.

Self-directed work and published projects demonstrate relevant practice. They are not paid-client delivery records or guarantees for this task.

Start with a short note

Three answers are enough. I will reply with what else is needed and whether a diagnosis comes first. You get a quote before any work starts.

Prepare a brief →

AI workflow & automation

Start with a short note

Three answers are enough. I will reply with what else is needed and whether a diagnosis comes first. You get a quote before any work starts.

What to send
  • Trigger, input, destination systems and desired outcome
  • Rules, failure paths and human approval steps
  • API access, sandbox, run frequency and cost ceiling
Illustration · hypothetical

Illustrative example — not a completed client order

The request

Classify selected emails into CSV, with mandatory human approval before any outbound send.

Example inputs

  • Trigger and filters, source fields and CSV schema
  • Sandbox/API availability and explicit access boundaries

To start, it helps to mention

  • What triggers the workflow, and what are its source and destination?
  • Which steps send or modify data and require human approval?

Confirm before work begins: Authorized APIs/test accounts, rate/cost caps and approval/operating ownership

Public/de-identified information only. Private access requires separate authorization.

Open email draft →

Preview the text
Services brief — A03 AI workflow & automation

First contact does not require a complete specification. Unknown fields may be blank or marked unknown.

What do you want to end up with?


What do you already have?


When do you need it? Any budget range?

Select and copy the brief manually if scripting or clipboard access is unavailable.

Public email is a contact starting point. This site has no booking, payment or file-upload service. zhenglimin.design@gmail.com

← Back to category
Prepare a brief →