Analyze Manual Text Evidence.prompt

# 35 - Analyze Manual Text Evidence Against Source Artifacts

## Purpose

Analyze a manually pasted text snippet that the user confirms is copied from the BRD or a trusted requirement source.

The goal is to determine whether the pasted text confirms, clarifies, conflicts with, or adds missing detail to the existing source-bound artifacts from steps 10–34.

This task may produce requirement/rule/update recommendations, but it must not directly rewrite the Functional Design, Detailed Design, code design, or implementation tasks.

This prompt is intended for focused text evidence review, especially when a user manually copies a small command syntax, table row, note, paragraph, or label from the source document after step 34.

## Operating rules

- Treat the pasted text as focused supplemental evidence.
- Do not analyze screenshots or images in this prompt.
- Do not infer hidden context beyond the pasted text and existing source artifacts.
- Do not invent requirements, business rules, validations, command behavior, API behavior, or design decisions.
- Preserve command names, option names, parameters, field names, product terms, acronyms, and code-like identifiers exactly.
- Follow `output/21_glossary.md` and `output/11_translation_policy.md`.
- Respect strikethrough rules from the pipeline:
  - Strikethrough-only content must not be treated as active behavior unless the source explicitly says otherwise.
  - If the pasted text came from strikethrough content, mark it as inactive/deprecated candidate evidence.
- Use pasted text to check existing requirements and rules, not to bypass the source-bound pipeline.
- Write the output in English.

## Tasks

### Precondition

Use the latest available evidence gate:

- If `output/34_FEATURE_UNDERSTANDING_REVIEW.md` exists, use it as the latest understanding gate.
- If the step 34 decision is `No-Go`, do not use the pasted text to bypass the review result.
- If the step 34 decision is `Fix required`, analyze the pasted text only as supplemental evidence and state that downstream FD/DD work should wait until the brief is fixed.
- If the step 34 decision is `Go` or `Go with minor notes`, continue normally.

Also verify whether the pasted text can be found or cross-checked in one of these artifacts:

- `output/12_normalized_evidence.md`
- `output/30_requirement_inventory.md`
- `output/31_business_rule_catalog.md`
- `output/32_open_questions.md`
- supporting raw extracted source, only if needed

### Instructions

Given the pasted text snippet:

1. Extract visible factual elements from the pasted text:
   - command names
   - operation names
   - options
   - parameters
   - required/optional indicators
   - constraints
   - unsupported cases
   - state names
   - input/output data
   - notes

2. Classify the snippet:
   - command syntax
   - business rule
   - operation constraint
   - data definition
   - UI/API behavior
   - state transition
   - note/comment
   - unknown

3. Map the snippet to existing source artifacts:
   - `output/30_requirement_inventory.md`
   - `output/31_business_rule_catalog.md`
   - `output/32_open_questions.md`
   - `output/21_glossary.md`
   - `output/12_normalized_evidence.md`

4. Determine whether the snippet:
   - confirms an existing requirement
   - clarifies an existing requirement
   - confirms an existing business rule
   - clarifies an existing business rule
   - adds a missing requirement candidate
   - adds a missing business rule candidate
   - creates or resolves an open question
   - conflicts with existing evidence
   - is duplicate/reference-only evidence
   - is too narrow to use without more context

5. If the pasted text contains enough evidence, propose requirement/rule candidates.
   Do not mark them as final unless they are already supported by existing source artifacts.

6. Recommend what to update:
   - requirement inventory
   - business rule catalog
   - open questions
   - glossary
   - feature understanding brief
   - no update required

## Inputs

### Manual input

The user will provide a pasted text snippet in chat.

Example format: