skills/101-index/internal/estimate-management/SKILL.md

---
name: estimate-management
description: Use when an internal 101 goal must find, read, create, or edit one or many estimates through the existing event API, including sourced or manual positions and their dates.
version: "1.0.0"
role: primary
invocation: internal
intents:
  - estimates.list
  - estimates.read
  - estimates.create
  - estimates.edit
depends_on:
  - entity-resolution
  - write-preflight
  - event-positions
required_tools:
  - whoami
  - list_events
  - get_event
  - create_estimate
  - edit_estimate
optional_tools:
  - list_projects
  - list_bills
  - list_contractors
  - upload_files
resources:
  - path: ../shared-resources/context-and-identity.md
    kind: semantic-guide
    required: true
  - path: ../shared-resources/events-and-positions.md
    kind: semantic-guide
    required: true
  - path: ../shared-resources/safety-and-permissions.md
    kind: semantic-guide
    required: true
completion:
  statuses:
    - готово
    - частично
    - заблокировано
---

# Estimate Management

Run only as the primary skill selected by `101-index`. An estimate uses existing estimate tools and their original API contract; never create another event type or MCP-only payload.

## Search and read

Resolve the company, project, and other named entities through `entity-resolution`. Use `list_events` with server filters and pagination for a list and `get_event` for details. Never treat the first page as the complete set or expose GUIDs without an explicit request.

## One record

For creation, prepare positions through `event-positions`, preserve their sources, `startDate`, and `endDate`, pass the exact create payload to `write-preflight`, and call `create_estimate`.

For an edit, first call `get_event`. Convert a targeted position change through `event-positions` into the complete resulting list while preserving neighboring rows and dates. After a fresh `write-preflight`, call `edit_estimate`. Stop and show a conflict on a concurrent change.

## Estimate series

A bulk request is a sequence of individual `create_estimate` calls, not a new batch payload.

1. Build the full queue and run a common preflight covering required fields, entities, machine schema, positions, and the ability to build every exact payload.
2. If any error is known in advance, create nothing; request all corrections in one message, then recheck the entire queue.
3. After the common check succeeds, briefly state the queue and start without another confirmation unless central policy requires one.
4. Repeat a fresh `write-preflight` before each item and make one API call.
5. A local data error marks that item not created but does not block later independent items.
6. A system error, lost permission, contract change, or uncertain write outcome stops the unprocessed tail.

Preserve an API receipt for every successful call. On continuation, skip proven created items, recheck proven not-started items, and first compare server state for an unknown outcome. Never retry an unknown write automatically.

Finish with `готово`, `частично`, or `заблокировано`; list created, not created, and not started estimates by human name, show fresh data, and give the smallest next step.