skills/learn-with-vera/references/get-started.md

# Get started with Vera

Use this short introduction when someone asks to get started, understand how
Vera works in general, or sends the catalogue's introductory request. Read the
localized outline in `get-started.json`. This is an introduction to using Vera,
not another professional workflow and not the older three/four-course onboarding
programme. It is optional and repeatable. A specifically requested specialist
lesson still starts directly.

Aim for 5–8 minutes of explanation and a short practice; actual execution and
questions can add time. Teach the general interaction through **one** existing
prepared case. Do not invent a new pipeline, recreate the teaching materials,
deliver a feature tour or require three/four completed courses to finish.

## 1. Choose something useful

Ask naturally what the professional does and what they would like to try. Reuse
what they already told you; one useful follow-up is enough when the goal is clear.
An existing user's current interests take priority over an old profile. Skip
the interview if requested. Do not collect client identifiers or credentials.

Read the installed Vera workflow catalogue and `local_courses.py list`. Suggest
3–4 distinct relevant operational workflows with supported kits in the user's
language, each with its plain name and concrete output. Apply the specialist's
jurisdiction and input scope; language alone does not establish jurisdiction.
Choose one with the learner. If they have no preference, offer the prepared
bank-reconciliation example: compare a fictional bank statement and ledger and
open the resulting reconciliation. Read its current skill and teacher notes.
The suggestions are next-course options, not a compulsory programme.

## 2. Ask and supply the files

Follow the parent teaching skill's host contract. In Codex use a repeatable
single-course session (`local_teaching.py begin`) for the chosen real workflow;
do not enroll this short introduction in the multi-course onboarding helper.
Use the existing teacher/working pair. Explain their roles when opening them:
questions and explanations here, execution and actual results in the working
chat. Follow the normal voice controls and visibility checks.

In Cowork follow the written single-conversation contract and its `prepare`
command. Never run desktop profile/session commands there or ask for a second
chat or voice. The reference's Codex instructions do not override that contract.

Materialize the selected kit in a fresh lesson directory using the normal host
helper. Open the supplied source files and identify what each contributes. Show
one ordinary request: desired result, files and relevant period or question.
Ask the learner what they want to obtain; pause for their answer. They do not
need to type internal workflow names or technical commands.

## 3. Execute and read the result

Run the actual current workflow with those prepared inputs, retaining its normal
reviews, managed runtime and delivery requirements. Explain questions Vera asks
as they arise: what is missing, which choice changes the result and who decides.
Do not narrate invented progress. If execution is blocked, keep it pending.

Open the actual generated result with the parent skill's preview procedure;
HTML must be rendered in a browser, never presented as source text. Show where
the file is saved and how to reopen it. Inspect one useful figure or finding,
trace it to an input and explain what still needs professional review. Ask the
learner to point out the result they would check first. A prepared specimen is
not today's output; a successful script is not professional approval.

## 4. Change something and try again

Use the selected kit's authored practice variation. Ask the learner to make its
small change as a natural request (for example, use the updated period or add
the supplied document). Wait for their request before executing. Run a fresh
attempt, preserve the first result, then open both and explain what changed.
If the learner asks for another variation, use only supported, checked fictional
inputs. Never manufacture their answer, confirmation or completed practice.

## 5. Leave with a next step

Ask the learner how they would ask Vera to do a similar task and which files they
would supply. Explain any remaining confusion. Give a brief recap containing
their useful request, the real demo/practice links, where to reopen the work and
the next course they chose from the suggestions. Do not automatically start it.
They can skip practice, leave, resume or use ordinary Vera work at any point;
describe skipped or unfinished steps accurately.

Record the real workflow's progress through the existing host mechanism. Keep a
short `getting-started-progress.md` alongside that lesson's files: introductory
step reached, suggested workflows, selected workflow/language, real output links,
the learner's actual answers and next step. Resume that record after verifying
the files and current skill. Do not reset other lessons or mark the separate
multi-course onboarding programme complete. A confirmed saved Codex profile may
inform teaching; this introduction requires no new account-wide profile. Cowork
uses only its connected lesson folder and does not claim access to Codex's state.

Do not send these notes or feedback to Mparanza. Retain the normal local tutorial
marker and professional-action boundaries. The selected host model processes
the conversation and any files it reads; saved lesson files do not mean offline
model processing. Follow the parent skill and selected workflow's data contract.