Skip to content
2 changes: 1 addition & 1 deletion skills/uipath-genome/references/execution-guide.md
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,7 @@ When the genome carries a design link ([genome-reading-guide.md § Design Links]

Gather facts the 1.3 scaffolding questions are built from before asking anything. Discovery is silent; it never replaces the questions.

1. Run the UiPath login status check (`uip login status --output json`); record what an expired session blocks (resource refresh, publish, deploy, Test Manager attachments — local init, validate, build, pack and run still work).
1. Run the UiPath login status check (`uip login status --output json`) and take the session's state from `Data.Status`, never from `ExpirationDate`, which is the current access token's; record what an expired session blocks (resource refresh, publish, deploy, Test Manager attachments — local init, validate, build, pack and run still work).
2. Confirm runtime host and edit surface; record which target frameworks and expression languages they support.
3. For every skill named in Build With / Components, run that skill's discovery commands for installed packages; record activity packages and versions available to the authoring host. When a project already exists in the build location, also record its declared dependencies and versions from its manifest.
4. Determine candidate build location: the folder containing the genome file, never the current working directory when that is a different repository.
Expand Down
6 changes: 3 additions & 3 deletions skills/uipath-genome/references/genome-reading-guide.md
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@ Rules:

## File Naming and Location

- Slug: split CamelCase on capitals, lowercase, hyphens for spaces, strip other characters. "Invoice Processing" → `invoice-processing`; "InvoiceIntake" → `invoice-intake`; "Email Triage & Routing" → `email-triage-routing`.
- Slug: split CamelCase on capitals, lowercase, hyphens for spaces and underscores, strip other characters. "Invoice Processing" → `invoice-processing`; "InvoiceIntake" → `invoice-intake`; "Email Triage & Routing" → `email-triage-routing`.
- A process genome's components are at `<process-slug>-genome/<component-slug>-genome.md`.

## Build With and Components
Expand Down Expand Up @@ -113,8 +113,8 @@ Extracted genomes only: where every part of the genome came from. A genome witho

1. **One row per Workflow step**; a test component's steps are its test cases.
2. **Source objects** are the objects whose behaviour the step implements, primary first: a rebuild that keeps the source's names names the step's file or workflow after it. A framework source's object is `` `Name` (id) `` or `` `Name` (id, {locator}) `` — the id identifies the copy; a UiPath source's is its project-relative path, a test case with its `testCaseId`.
3. **Data sets** are the data sets the step's own objects read, or `none`.
4. **Captures** is the number of screen captures the source takes in the step; the build checks the rebuild's evidence against it.
3. **Data sets** are the data sets that drive the step — the rows or items it is run over — or `none`; a configuration table, a file the step writes or a URL is in Notes.
4. **Captures** is the number of screenshots the source takes in the step, never a count of recorded UI actions; the build checks the rebuild's evidence against it.
5. **Notes** is free text no check reads.

**Unused parts.** A kept workflow nothing invokes is built like any other step. Parts in the `Not covered` or `Unused` row are left out of the build unless translated first.
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -90,7 +90,7 @@

### Flow 1 — {unit of work}: {producer} → {consumer}

**Role:** {producer | consumer | consumer of Flow {M} and producer of this flow} — {one sentence}. *(component of a process only)*
**Role:** {producer | consumer | producer and consumer | consumer of Flow {M} and producer of this flow} — {one sentence}. *(component of a process only)*

**Unit of work:** one {item}; reference {identifying field(s)}; fields as in {Interface / Handoffs row}; {volume and cadence}; chosen because {items fail, retry and are reported independently at this level; the reference exists; the retry cost is acceptable}.
**Alternative units of work:** one {other item} — fits when {condition}. *(every viable granularity, or "none")*
Expand Down
4 changes: 2 additions & 2 deletions skills/uipath-planner/references/author/author-guide.md
Original file line number Diff line number Diff line change
Expand Up @@ -84,12 +84,12 @@ Then add these capabilities even when the user did not name them:

| User mentions | Add |
|---|---|
| PDFs, scanned documents, forms | Document Understanding activities in `uipath-rpa` step; `uipath-ixp` component when custom extraction model needed |
| PDFs, scanned documents, forms | What [skill-mapping-guide.md § Decision Tree for Ambiguous Steps](../shared/skill-mapping-guide.md) item 3 names |
| Clicking, typing, reading a screen, desktop app | UI automation in `uipath-rpa` |
| REST APIs, connectors, SaaS systems | Integration Service connector activities (`uipath-rpa`) or `uipath-api-workflow` component when no UI involved |
| Many items processed independently, retries, resilience, several robots | Transactional Shape with the As-is, the split options and the evidence ([transactional-shape-guide.md](../shared/transactional-shape-guide.md)); an Orchestrator queue under Platform Dependencies only when the description itself shares items across robots or runs — otherwise the store follows the split chosen at execution, not imposed |
| Coordinating several automations, waiting on events | `uipath-maestro-flow` (short-lived) or `uipath-maestro-bpmn` (long-running, human lanes) coordinator component |
| Judgement, classification, summarisation, free-text decisions | `uipath-agents` component |
| Judgement, classification, summarisation, free-text decisions | What [skill-mapping-guide.md § Decision Tree for Ambiguous Steps](../shared/skill-mapping-guide.md) item 5 names |
| Human review, approval, sign-off | Human-in-the-loop checkpoint in coordinator; actor row in Actors and Systems |
| Screen for end users, dashboard | `uipath-coded-apps` component |
| Several automations or test suites drive the same application's screens | One `uipath-rpa` library component (screens and shared actions, Interface as per-workflow argument tables) plus consumer components; process genome states the library builds and packs first |
Expand Down
10 changes: 5 additions & 5 deletions skills/uipath-planner/references/extract/extract-guide.md
Original file line number Diff line number Diff line change
Expand Up @@ -98,7 +98,7 @@ Read every relevant file. No sampling, no "the helpers are similar".

**Values are read, not inferred from their rendering.** Run the source guide's renderer self-test first when it has one. Every fact of the kinds [extracted-genome-guide.md § Source values](../shared/extracted-genome-guide.md) lists is written as the source states it. A fact the rendering does not show is read from the source object before it is written; a fact that cannot be read is flagged `*[Inferred]*` and named in the Source Map, never filled with a plausible number.

**Wrappers are not components.** Source code that does what a UiPath driver or activity package does is the source's substitute for it: a library that drives a technology through an identity passed to it — find the element by id, then click, type, read — substitutes for a driver, and an action whose code does what one or a few activities do — open, read or write a workbook, send a mail, copy a file — substitutes for an activity package. Its calls are steps of the component that makes them, written as the behaviour its callers see (an error it swallows, a failure flag a caller tests), never as its code. A driver substitute's calls are UI steps on the element the identity names, and the identities are that component's UI targets (recorded in the Source Map's inventory with the recorder's captures). An action with logic of its own around the activities — a rule over the values, the format of a report it writes, a property no activity sets — is no wrapper, nor is an action on an application's own screens (that application's UI library), nor are a driver substitute's logon, session and certificate workflows: each stays the component the source guide's § Component Detection makes it; that section also names the framework's wrappers.
**Wrappers are not components.** Source code that does what a UiPath driver or activity package does is the source's substitute for it: a library that drives a technology through an identity passed to it — find the element by id, then click, type, read — substitutes for a driver, and an action whose code does what one or a few activities do — open, read or write a workbook, send a mail, copy a file — substitutes for an activity package. Its calls are steps of the component that makes them, written as the behaviour its callers see (an error it swallows, a failure flag a caller tests), never as its code. A driver substitute's calls are UI steps on the element the identity names, and the identities are that component's UI targets (recorded in the Source Map's inventory with the recorder's captures). An action with logic of its own around the activities — a rule over the values, the format of a report it writes, an object-model call no activity makes, for an item's property, the application's own state the caller relies on once the action returns, or one of its commands — is no wrapper (a setting the action sets and restores around its own work goes with its wrapper), nor is an action on an application's own screens (that application's UI library), nor are a driver substitute's logon, session and certificate workflows: each stays the component the source guide's § Component Detection makes it; that section also names the framework's wrappers.

**Related resources.** Read each resource from Step 1 after the source's signals. The source decides behaviour: a resource never adds, removes or changes a step, rule or value. It only settles what the source cannot show (purpose, actors, why a rule exists, volumes, success criteria), and a line it settles loses `*[Inferred]*`. List each resource in the Source Map's `Related resources` row with what it settled, or "not read" and why. Record every disagreement with the source in `Resource discrepancies`, never in the body.

Expand All @@ -120,17 +120,17 @@ A finding that the source misbehaves is a suspected defect until two passes sett
|---|---|---|---|
| Workflow files or nodes carrying logic | 1-3 | 4-8 | 9+ |
| Distinct target applications or connectors | 1-2 | 3-4 | 5+ |
| Entry points or triggers | 1 | 1-2 | 3+ |
| Entry points or triggers | 1 | 2 | 3+ |
| Conditional branches (if / switch / gateway / decision) | 0-3 | 4-10 | 11+ |
| Error-handling constructs | 0-1 | 2-3 | 4+ |
| Composite UI interactions carrying a contract (type-ahead pick, custom option list, segmented or picker date, menu path, row-by-content, dialog answer) | 0-1 | 2-4 | 5+ |
| Test cases of a test component, or checkpoints they assert (take the higher) | 1-3 cases / ≤6 checkpoints | 4-8 / 7-20 | 9+ / 21+ |
| Platform resources (queues, assets, buckets, connections) | 0 | 1-2 | 3+ |
| Human-in-the-loop waits | 0 | 0 | 1+ |
| Human-in-the-loop waits | 0 | — | 1+ |

A **test component** is rated by its test cases and the checkpoints they assert; a **library** by its public workflows, counted as its workflow files, with the library as one entry point. Neither takes the population matrix's step minimums ([genome-format-guide.md § Population Matrix](../shared/genome-format-guide.md)).

Count only the rows that apply to the component: the composite UI row is left out for a component that drives no UI, the test row for one that is not a test component. Take the level most counted rows land on; ties go lower. Process genomes: [genome-format-guide.md § Complexity](../shared/genome-format-guide.md).
Count only the rows that apply to the component: the composite UI row is left out for a component that drives no UI, the test row for one that is not a test component, the workflow row for one that is. Take the middle level of the counted rows, ordered simple to complex, the lower of the two middle ones when the count is even (2 simple, 1 medium, 2 complex → medium). Process genomes: [genome-format-guide.md § Complexity](../shared/genome-format-guide.md).

### Step 6 — Map signals to sections

Expand All @@ -156,7 +156,7 @@ Write the **process genome first** (when applicable), then each **component geno

### Step 6b — Complete the Source Map as the migration contract

Fill every row [extracted-genome-guide.md § Source Map](../shared/extracted-genome-guide.md) lists: framework, export root path and identity, the step table of every component genome (each step's source objects, the data sets its own objects read, its captures), checkpoints per step of a test component and of every library step a test component calls (source guide's § Evidence says which step kinds count), inventory counts. The call graph of Step 4 gives the source objects; a caller, and a callee that is a step of its own, go to the step's Notes. Nothing is written beside the genome.
Fill every row [extracted-genome-guide.md § Source Map](../shared/extracted-genome-guide.md) lists: framework, export root path and identity, the step table of every component genome (each step's source objects, data sets and captures), checkpoints per step of a test component and of every library step a test component calls (source guide's § Evidence says which step kinds count), inventory counts. The call graph of Step 4 gives the source objects; a caller, and a callee that is a step of its own, go to the step's Notes. Nothing is written beside the genome.

Check the Source Map before offering edits, in every genome the extraction wrote, against [genome-format-guide.md § Writing Rules](../shared/genome-format-guide.md) and against the source: every Workflow step has exactly one step-table row, and every source object and data set a row names exists in the source under that name. With a framework export, read the names from the source guide's inventory, run into a scratch folder **outside the genome's folder** (a temp directory, or one deleted after the check). Keep the inventory files the pack's `data` command wrote there until Step 7's strict check has run: that check resolves every step table against them. What is still open is named when edits are offered (Step 7).

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -14,8 +14,8 @@ Where every part of the genome came from: a table of contract rows — one row p

1. **One row per Workflow step**, keyed by its number exactly as the Workflow numbers it — never `Step 3`, a name or a range. A test component's steps are its test cases. A step built from the same objects as another repeats them.
2. **Source objects** lists only the objects whose behaviour the step implements, separated by `; `, the step's primary object first: the one whose name a rebuild that keeps the source's names gives the step's file or workflow. A framework source writes each as `` `Name` (id) `` or `` `Name` (id, {locator}) ``: names repeat across folders in most frameworks, so the id is what identifies the copy, and the locator is the position inside the object that the source guide's § Provenance names (lines, a page, nodes, test steps), backticked names allowed. A UiPath source writes the workflow's project-relative path, `` `Folder/Workflow.xaml` ``, a test case with its `testCaseId` in the parentheses ([uipath-source-guide.md § Provenance](../extract/uipath-source-guide.md)). A caller, and a callee that is a step of its own, go to Notes; a window or a control never appears here — it is the target catalog's.
3. **Data sets** lists the data sets the step's own objects read, written like a source object, or `none`: a test case's own data set, and one it borrows from another case or a shared set (login rows, interface data). A library step's cell is `none` unless the library reads a data set itself: what its callers pass is their data, on their own steps. The column gives the kind, so a root and its data set that share a name, or a data set whose id equals a process id, are never mistaken for one another.
4. **Captures** is the number of screen captures the source takes inside the step, `0` when it takes none; the build checks the rebuild's evidence against it.
3. **Data sets** lists the data sets that drive the step's own objects — the rows or items it is run over — written like a source object, or `none`: a test case's own data set, one it borrows from another case or a shared set (login rows, interface data), a work queue it takes items from. A set the framework's recordset list marks `drives: false` — a configuration table, a file the step writes, a URL — is provenance, named in Notes. A library step's cell is `none` unless the library reads a data set itself: what its callers pass is their data, on their own steps. The column gives the kind, so a root and its data set that share a name, or a data set whose id equals a process id, are never mistaken for one another.
4. **Captures** is the number of screenshots the source takes inside the step, `0` when it takes none — never a count of recorded UI actions; the build checks the rebuild's evidence against it.
5. **Notes** is free text — callers, callees, lines, findings, quirks — and the only cell no check reads.

**Unused parts** — enabled objects no entry point reaches ([extract-guide.md](../extract/extract-guide.md) Step 2b) — are written as the Unused parts answer says, except an unused action that only works a technology, with no business rule of its own, which is left out under every answer. Extracted, each is written like any other part: an uncalled library workflow is a public workflow of its library, an uncalled entry point or test case a step of its component, and a workflow nothing invokes a step of its component, after the steps the entry points reach, whose first sentence says that no step invokes it today; its step-table row's Notes say `no caller in the export`. Left out, the live surface goes to the `Not covered` row and the rest to an `Unused` row, each by name and id with what it does in a few words, so a partial re-extraction can add it later ([extract-guide.md § Partial Re-extraction](../extract/extract-guide.md)).
Expand Down
Loading
Loading