Skip to content

Arc Skills / Requirements

Clean and map a requirements spreadsheet

Use Arc Skills to turn a requirements spreadsheet into a visible field mapping and reviewed import proposal. Preserve original IDs, wording and metadata, reconcile every source row, and show engineering issues separately from mapping decisions.

Use this skill

Use Arc Skills to clean and map my requirements spreadsheet.

Inputs:
- The source workbook or CSV, sheet and revision
- Target fields and any identifier or blank-cell conventions

Return the original requirements, column mapping, transformation log, duplicate-ID findings, count reconciliation and proposed parent links. Keep engineering-quality findings separate from mapping warnings.

Preserve source IDs, wording and unmapped columns. Export new files without overwriting inputs.

Set up the toolkit · Read the skill instructions

What you provide and what you get

Inputs and outputs
What you haveHow it is usedWhat you get
Requirements spreadsheet and sheet identityInspect rows and column conventionsA source-preserving requirement table
Target fields and relationship conventionsMake every column disposition explicitField mapping and proposed links
Raw source and export destinationReconcile records and preserve provenanceA new import proposal plus separate engineering findings

Start with four requirements

Synthetic input from the toolkit’s public example. It has six source columns, four requirement records and one parent reference. Cleanup means mapping and preserving the data; engineering changes remain separate proposals.

Original requirements
Source IDOriginal statementParent source ID
REQ-001During imaging the payload shall deliver calibrated frames at a rate of at least 2 Hz.Blank
REQ-002During imaging the processor shall produce each calibrated frame within 400 ms.REQ-001
REQ-003The controller shall report a detected sensor fault within 100 ms.Blank
REQ-004The unit shall provide an appropriate warning quickly.Blank
Download the source CSV

Make every column’s disposition explicit

Field mapping
Source columnDestinationTreatment
IDSource/import IDPreserve the source identifier as text. Local proposal IDs are separate.
NameNamePreserve the source name.
StatementRequirement statementPreserve the original wording, including vague language.
Parent IDProposed derivation relationshipResolve REQ-002 to REQ-001; preserve blank cells as no proposed parent.
NotesNotesRetain context without treating it as an approved requirement.
Customer fieldRetained source metadataNo owner field is inferred. Keep the unmapped column available for review.

Reconcile the output before reviewing it

The local import tool produced 4 source records → 4 candidate requirements; 0 empty records, 0 rejected records, 1 proposed parent link and 0 unresolved links. The proposal passed exchange-schema validation. That validates its structure, not requirement quality or the correctness of its derivation.

Engineering review after mapping
RequirementIssue or open decisionNext action
REQ-001Frame delivery is quantified at ≥2 Hz, but measurement interval and operating configuration are not supplied.Define the interval, modes and evidence needed to decide the rate.
REQ-002The 400 ms processing limit is linked to frame delivery; that alone does not prove complete parent coverage.Review remaining latency, pipeline and scheduling assumptions before accepting the derivation.
REQ-003The note refers to an interface definition of detection, which is not included.Inspect that definition before deciding timing coverage.
REQ-004“Appropriate” and “quickly” do not identify the warning behaviour or deadline.Request the warning, trigger, recipient and latency. Preserve the original statement.

Map the source before proposing edits

  1. Inspect the actual sheet, row boundaries and column semantics before selecting a mapping.
  2. Preserve identifiers as text, retain original statements and give every source column a disposition.
  3. Resolve relationship targets and distinguish duplicate identity conflicts from repeated wording.
  4. Reconcile source, candidate, rejected and unresolved records before exporting a new proposal.

Sources and further reading

Synthetic public example. The import proposal was generated on 3 October 2026 with the local file tool; the mapped CSV and issue table are authored review aids.