smaqit-requirements-extract

name	smaqit.requirements-extract
description	Use this skill when the user wants to extract structured requirements candidates from raw project assets such as code prototypes, documents, or meeting notes. Produces `.smaqit/requirements-extract.md` — an enumerated, categorised inventory that specification agents can consume directly. Also use when the user asks to analyse assets, mine files for requirements, or prepare input for any spec workflow.
metadata	{"version":"1.0.1"}

Requirements Extraction

Steps

1. Determine the asset path. Use assets/raw/ unless the user specifies another path. If the directory is absent or empty, ask for the path before proceeding.

2. Read every file in the directory. If any file has a .pdf extension, invoke smaqit.utils.read-pdf first and work from the sidecar .extracted.txt file it produces.

3. Scan each file in full — including code logic, not just comments and data declarations — and extract items into these six fixed categories:

   Category	What to capture
   Data models	Entity shapes, field names, types, keys
   Business formulas	Scoring algorithms, calculation rules, thresholds (e.g. attrition risk tiers, score normalisation)
   Navigation / UI flows	Screen names, tab lists, routing logic, onboarding gates
   State machine rules	Locking conditions, completion triggers, immutability constraints
   Integration contracts	API shapes, response keys, 3rd-party SDKs referenced
   Constraints and anomalies	Range discrepancies, deviations from standard patterns, explicit exclusions

   For each item, record a source annotation: filename + line number or paragraph reference. Do not normalise conflicting values — preserve discrepancies exactly as found.

4. Write the inventory to .smaqit/requirements-extract.md using this structure:

   • One ## heading per category (all six must be present, in the order above)
   • Bullet items under each heading; each bullet ends with its source annotation in parentheses
   • Write None identified. under any empty category — do not omit the heading
   • ## Ambiguities section at the end listing every conflicting value, undefined term, or unclear constraint found; each entry states what conflicts and which source files are involved

5. Report to the user: item count per category and total ambiguity count. If the ambiguity count exceeds 5, name the three most impactful and recommend resolving them before spec generation.

Output

.smaqit/requirements-extract.md — structured inventory with six categorised sections and an ## Ambiguities section.

Scope

• Does NOT write specifications. Output is raw material for specification agents (smaqit.business, smaqit.functional, etc.).
• Does NOT infer requirements absent from the source files.
• Does NOT process PDFs directly — invoke smaqit.utils.read-pdf first.

Examples

Input: User places code.txt (React prototype, ~1500 lines) in assets/raw/ and invokes /requirements.extract.

Output: .smaqit/requirements-extract.md containing score formulas, 7 tab names, habit data model, baseline locking trigger, and 2 flagged range discrepancies across 6 categorised sections.

Gotchas

• Source files may use inconsistent ranges for overlapping concepts (e.g. one instrument scale using a 0–4 range while another uses 1–5). Annotate range values with their source file and do not normalise — preserve the discrepancy as-is and flag it in ## Ambiguities.
• Code files (e.g. React components) embed formulas and state machine rules inside component logic, not just in data declarations or comments. Scan the full file body.
• The extraction output is not a specification. Functional specification agents resolve ambiguities — this skill only surfaces them.

Completion

• All files in the asset directory read (PDF files processed via smaqit.utils.read-pdf first)
• .smaqit/requirements-extract.md written
• All six category headings present (empty categories contain "None identified.")
• ## Ambiguities section present
• Report delivered: item count per category + ambiguity count

Failure Handling

Situation	Action
Required input not provided	Request the missing information before proceeding
Gathered input is ambiguous	Flag the ambiguity and ask for clarification
Subagent invocation fails	Report the failure with context; do not silently retry
Output artifact already exists	Confirm with user before overwriting
assets/raw/ is absent or empty	Ask the user for the asset path before proceeding
File is a PDF	Invoke smaqit.utils.read-pdf first, then continue with the sidecar
Ambiguity count exceeds 5	Summarise the top 3 most impactful in the report; recommend resolving before spec generation