Skip to content

Get the Seed

The seed is one public repository holding everything the exercise gives you: the requirements, the environment reference, the contracts, and one prompt per block. Every path in it is relative to its root, so the folder you create now is where the whole day happens.

Clone it

git clone https://github.com/uipath-practice/PropertyClaimsSeeds.git ClaimCase-<seat>
cd ClaimCase-<seat>

Name the folder for your seat. If you cannot use git:

curl -L https://github.com/uipath-practice/PropertyClaimsSeeds/archive/refs/heads/main.zip -o seed.zip
unzip seed.zip 
mv PropertyClaimsSeeds-main ClaimCase-<seat>

Prefer the clone. With git, git status answers what did I build and git diff answers what did I change, so you write up your build from evidence rather than from memory.

What's inside

Path What it is
README.md The front door: the block sequence and how to work it
PDD.md The business requirements: the document the whole build works from
CONFIG.md The environment reference card: tenant, names, what already exists
PROGRESS.md A skeleton your build fills in: the memory between blocks
contracts/ The pinned interfaces: the claim record schema, the component list, the review-task shape, the provided automations
method/ The delivery method the blocks follow, and the PDD and SDD guides: reference only, not used during the build.
known-issues/ Platform behaviours worth knowing; they save your agent time.
1-design/ … 6-process-app/ One folder per block: prompt.md, cookbook.md, and sometimes a gate script or a helper (4-verify/agent_inputs.py fetches the inputs the case sent an agent).
log-finding.py How build findings get logged 
ClaimCase-Pro % ls -la 
total 352 
drwxr-xr-x@ 26    832    4 Sep 14:17  . 
drwxr-xr-x 9      288    4 Sep 14:16  .. 
drwxr-xr-x@ 3     96     4 Sep 14:17  .claude 
drwxr-xr-x@ 12    384    4 Sep 14:16  .git 
drwxr-xr-x@ 3     96     4 Sep 14:17  .workshop 
drwxr-xr-x@ 5     160    4 Sep 14:16  1-design 
drwxr-xr-x@ 4     128    4 Sep 14:16  2-plan 
drwxr-xr-x@ 7     224    4 Sep 14:16  3a-extraction 
drwxr-xr-x@ 4     128    4 Sep 14:16  3b-entity 
drwxr-xr-x@ 4     128    4 Sep 14:16  3c-agents 
drwxr-xr-x@ 5     160    4 Sep 14:16  3d-case 
drwxr-xr-x@ 4     128    4 Sep 14:16  3e-run 
drwxr-xr-x@ 6     192    4 Sep 14:16  3f-validation 
drwxr-xr-x@ 4     128    4 Sep 14:16  4-verify 
drwxr-xr-x@ 4     128    4 Sep 14:16  5-ship 
-rw-r--r--@ 1     15419  4 Sep 14:16  AGENTS.md 
-rw-r--r--@ 1     225    4 Sep 14:16  CLAUDE.md 
-rw-r--r--@ 1     12449  4 Sep 14:16  CONFIG.md 
drwxr-xr-x@ 6     192    4 Sep 14:16  contracts 
drwxr-xr-x@ 5     160    4 Sep 14:16  known-issues 
-rwxr-xr-x@ 1     23540  4 Sep 14:16  log-finding.py 
drwxr-xr-x@ 7     224    4 Sep 14:16  method 
-rw-r--r--@ 1     98226  4 Sep 14:16  PDD.md 
-rw-r--r--@ 1     1154   4 Sep 14:16  PROGRESS.md 
-rw-r--r--@ 1     9818   4 Sep 14:16  README.md 
-rw-r--r--@ 1     38     4 Sep 14:16  VERSION

Provided vs. built

Your Orchestrator seat folder comes with a few dependencies: mechanical RPA automations, storage buckets and a reusable IXP project for claims. The tenant also exposes shared resources. Bind those rather than building your own, so your time goes into the parts that need thinking, not basic scripts.

What you will produce

  • sdd.md and tasks.md: the plan (block 1 & 2)
  • Build/ClaimCase-<seat>/: the solution, holding the Maestro case and the seven Agents
  • Build/claim-review-<seat>/: the standalone Coded Action App
  • a filled PROGRESS.md, your logged findings, and at the end the operator's runbook.md

Pre-flight

  • uip login status: right tenant, right folder. Nearly every "not found" is a wrong tenant, a wrong folder or a stale cache, so build the habit now
  • uip --version: the CLI and each tool update themselves on their first use each day, within the version line CONFIG.md pins, so the version you ran can change between blocks. Note it at each block start.
  • python3 log-finding.py --identify: declares your agent, model and CLI version once, so results are diagnosable
  • Open README.md in the seed and read "The sequence"; it takes five minutes and explains the whole day