📄 Per-Document Methodology (Atomic Evidence Layer)

📊 Family E — Deep Analysis of Individual European Parliament Documents
🎯 Document Interrogation · Evidence Grading · Clause-Level Analysis · Provenance Mapping

📋 Document Owner: CEO | 📄 Version: 1.1 | 📅 Last Updated: 2026-04-25 (UTC) 🔄 Review Cycle: Quarterly | ⏰ Next Review: 2026-07-31 🏢 Owner: Hack23 AB (Org.nr 5595347807) | 🏷️ Classification: Public

---

🔄 Tradecraft Anchors

Element	Value	Reference
F3EAD Stage	EXPLOIT — extract actionable data from raw documents	Document-level analysis extracts structured facts from unstructured parliamentary text
PIRs Served	Each per-document file serves PIRs by extracting facts that answer specific intelligence requirements	See political-style-guide.md §PIR/EEI
Admiralty Floor	Source grade derives from document provenance: plenary = A1, committee = A2, press = B3	See political-style-guide.md §Admiralty
WEP Requirement	Forward-looking claims in per-document analysis use WEP bands	See political-style-guide.md §WEP + ODNI
ICD 203 Gate	Standard 1 (clear source), Standard 5 (distinguish facts from assumptions)	See political-style-guide.md §ICD 203
SAT(s)	Content Analysis, Chronological Review, SWOT Framing, Stakeholder Identification	See political-style-guide.md §SATs

---

🎯 Purpose

Family E delivers atomic evidence — the indivisible facts extracted from individual European Parliament documents. While Family A synthesizes across documents, Family E goes deep into one document at a time, producing:

1. per-file-analysis/{doc-id}.md — comprehensive clause-by-clause analysis of each document
2. document-analysis-index.md — summary index linking all per-document analyses

When Family E is Produced

Family E analysis is triggered whenever the workflow retrieves full document content via:

• get_adopted_texts (adopted resolutions, legislative positions)
• get_committee_documents (committee reports, opinions)
• get_plenary_documents (motions, amendments)
• get_procedures (procedure dossier documents)
• get_external_documents (Commission proposals, Council positions)
• search_documents (keyword-matched document retrieval)

Rule: Metadata-only retrieval (titles, dates, reference numbers) does not qualify as evidence for analysis. Per-document analysis requires full document content download.

---

📊 Document Type Taxonomy (European Parliament)

%%{init: {"theme":"dark","themeVariables":{"primaryColor":"#1565C0","primaryTextColor":"#ffffff","primaryBorderColor":"#0A3F7F","lineColor":"#90CAF9","secondaryColor":"#2E7D32","secondaryTextColor":"#ffffff","tertiaryColor":"#FF9800","tertiaryTextColor":"#000000","mainBkg":"#1565C0","secondBkg":"#2E7D32","tertiaryBkg":"#FF9800","noteBkgColor":"#FFC107","noteTextColor":"#000000","errorBkgColor":"#D32F2F","fontFamily":"Inter, Helvetica, Arial, sans-serif"}}}%%
flowchart TB
    classDef leg fill:#1565C0,stroke:#0D47A1,color:#FFFFFF
    classDef com fill:#4CAF50,stroke:#1B5E20,color:#FFFFFF
    classDef parl fill:#FF9800,stroke:#E65100,color:#FFFFFF
    classDef ext fill:#7B1FA2,stroke:#4A148C,color:#FFFFFF
    classDef qa fill:#D32F2F,stroke:#B71C1C,color:#FFFFFF

    subgraph LEG["Legislative Documents"]
        A1[COM — Commission Proposal<br/>Legislative initiative origin]:::leg
        A2[A* — Committee Report<br/>Committee recommendation]:::leg
        A3[RC — Resolution Compromise<br/>Cross-group negotiated text]:::leg
        A4[TA — Adopted Text<br/>Final plenary decision]:::leg
    end

    subgraph COM["Committee Documents"]
        B1[IMCO-AD — Opinion<br/>Advisory to lead committee]:::com
        B2[ENVI-PR — Draft Report<br/>Lead committee draft]:::com
        B3[AGRI-AM — Amendments<br/>Committee amendments]:::com
        B4[PE* — Working Document<br/>Preparatory material]:::com
    end

    subgraph PARL["Parliamentary Activities"]
        C1[MOT — Motion for Resolution<br/>Political statement]:::parl
        C2[QO — Oral Question<br/>Direct scrutiny]:::parl
        C3[QE — Written Question<br/>Documentary scrutiny]:::parl
        C4[CRE — Verbatim Report<br/>Plenary transcript]:::parl
    end

    subgraph EXT["External Documents"]
        D1[Council Position<br/>Council first/second reading]:::ext
        D2[Trilogue Document<br/>Interinstitutional negotiation]:::ext
        D3[ECJ Ruling<br/>Court of Justice decision]:::ext
    end

    A1 --> A2
    A2 --> A3
    A3 --> A4
    B2 --> A2
    B1 --> B2
    B3 --> B2

---

🗂️ Document-Type Analysis Templates

EP Document Types and Analysis Requirements

Document Type	Code	MCP Tool	Required Analysis Sections	Min Lines
Commission Proposal	COM(20XX)NNN	get_external_documents	Problem statement, legal base, impact assessment summary, stakeholder positions	200
Committee Report	A*-0NNN/20XX	get_committee_documents	Rapporteur recommendation, key amendments, vote distribution, committee splits	180
Adopted Text	P*_TA(20XX)NNNN	get_adopted_texts	Final position, material changes from committee, implementation timeline	150
Motion for Resolution	B*-0NNN/20XX	get_plenary_documents	Signatory analysis, political-group alignment, precedent context	140
Compromise Amendment	RC-B*-0NNN/20XX	get_plenary_documents	Cross-group negotiation trace, substantive trade-offs, abstention patterns	160
Written Question	QE-NNNNNN	get_parliamentary_questions	Question framing, Commission response, follow-up indicators	100
Oral Question	QO-NNNNNN	get_parliamentary_questions	Debate context, supplementary questions, Minister/Commissioner response	120
Committee Opinion	XX-AD-0NNN/20XX	get_committee_documents	Advisory scope, divergence from lead committee, voting pattern	130
Legislative Procedure	20XX/NNNN(COD/NLE)	get_procedures	Stage progression, timeline analysis, blocking points	180
Roll-Call Vote	A*-NNNN/20XX-VOTE	get_voting_records	Vote distribution, political-group cohesion, notable defections	150

---

📋 Per-Document Analysis Template Structure

Each per-file-analysis/{doc-id}.md file follows this canonical structure:

1. Document Header Block

# Per-Document Analysis: {DOC-ID}

**Document Type:** {Commission Proposal / Committee Report / Adopted Text / ...}
**Reference:** {Full EP reference number}
**Date:** {Publication date, ISO format}
**Source Grade:** {Admiralty code, e.g., A1 — Primary EP plenary record}
**Procedure:** {Linked procedure ID, e.g., 2024/0001(COD)}

---

2. Executive Summary (≤100 words)

BLUF (Bottom Line Up Front) statement of the document's significance, following ICD 203 Standard 2 format.

3. Document Classification

Dimension	Classification	Evidence
Political Significance	P0 / P1 / P2 / P3	{1-sentence justification}
Institutional Impact	HIGH / MEDIUM / LOW	{1-sentence justification}
Policy Area	{Primary Europarl classification}	{DG/Committee assignment}
Geographic Scope	EU-wide / Regional / MS-specific	{Named member states affected}
Temporal Horizon	Immediate / Short-term / Long-term	{Timeline markers from text}

4. Clause-Level Analysis

For legislative and procedural documents, provide clause-by-clause analysis:

Article/Clause	Summary	Political Implications	Stakeholder Impact
Art. 1	{Core provision summary}	{Who wins/loses}	{Affected actors}
Art. 2	...	...	...
Recital N	{Narrative framing}	{Political intent signals}	{Policy context}

5. Actor Mapping

Actor	Stated Position	Evidence	Inferred Interest
{MEP / Group / MS / Institution}	Support / Oppose / Neutral	{Citation to text}	{Strategic motivation}

6. Political Group Analysis

Political Group	Official Position	Key MEP Voices	Coherence Score
EPP	{Position}	{Named MEPs}	{HIGH/MED/LOW}
S&D	...	...	...
Renew	...	...	...
Greens/EFA	...	...	...
ECR	...	...	...
ID/PfE	...	...	...
The Left	...	...	...
NI	...	...	...

7. Procedural Analysis

%%{init: {"theme":"dark"}}%%
flowchart LR
    classDef complete fill:#4CAF50,stroke:#1B5E20,color:#FFFFFF
    classDef current fill:#1565C0,stroke:#0D47A1,color:#FFFFFF
    classDef pending fill:#9E9E9E,stroke:#424242,color:#FFFFFF

    A[Commission Proposal<br/>2024-01-15]:::complete
    B[Committee Reference<br/>2024-02-01]:::complete
    C[Committee Vote<br/>2024-05-20]:::current
    D[Plenary Vote<br/>TBD]:::pending
    E[Council Position<br/>TBD]:::pending
    F[Trilogue<br/>TBD]:::pending
    G[Final Adoption<br/>TBD]:::pending

    A --> B --> C --> D --> E --> F --> G

8. Cross-Reference Section

Related Document	Relationship	Relevance
{Reference}	{Amends / Supersedes / Responds to / References}	{Why relevant}

9. Forward Indicators

Indicator	Target Date	Monitoring Source
{What to watch for}	{ISO date or date range}	{EP API endpoint / Committee schedule / Plenary calendar}

10. Source Metadata

Field	Value
EP API Endpoint	{Specific API call used}
Retrieval Timestamp	{ISO timestamp}
Content Hash	{SHA-256 of full text, for provenance}
Word Count	{Full document word count}

---

🔍 Evidence Grading for Per-Document Analysis

Admiralty Grades for EP Document Sources

Grade	Source Type	Example
A1	Primary plenary record, official vote	Roll-call vote record from EP plenary
A2	Official committee document, published position	ENVI committee report (PE-document)
B1	EP press release, official spokesperson	EP News item, President's statement
B2	MEP official statement, parliamentary question	Written question + answer
B3	MS government position via Consilium	Council position document
C2	Established press with named sources	Politico/EUobserver with MEP quotes
C3	Established press with anonymous sources	Brussels bureau analysis piece
D3	Unverified social media, rumor	MEP tweet without corroboration

Content Quality Assessment

Each per-document analysis must pass these quality gates:

• Full document content retrieved (not metadata-only)
• BLUF summary ≤100 words, follows ICD 203 format
• Classification table complete (all 5 dimensions)
• Actor mapping identifies ≥4 distinct stakeholders
• Political group analysis covers all 8 categories
• Forward indicators identify ≥3 monitoring points
• Source metadata complete with retrieval timestamp

---

📊 Document Analysis Index Structure

The document-analysis-index.md file provides a navigable summary:

# Document Analysis Index

**Run Date:** {ISO date}
**Total Documents Analyzed:** {N}
**Source APIs:** {get_adopted_texts, get_committee_documents, ...}

## Summary Statistics

| Document Type | Count | High-Significance | Medium | Low |
|---------------|-------|-------------------|--------|-----|
| Adopted Texts | N | N | N | N |
| Committee Reports | N | N | N | N |
| Motions | N | N | N | N |
| Questions | N | N | N | N |

## Document List

| Doc ID | Type | Significance | Key Finding | Link |
|--------|------|--------------|-------------|------|
| TA(2024)0123 | Adopted Text | P0 | {1-line summary} | [Analysis](per-file-analysis/TA-2024-0123.md) |
| ... | ... | ... | ... | ... |

## Cross-Reference Network

{Mermaid graph showing document relationships}

---

🔗 Integration with Other Methodology Families

Upstream Data Sources (Stage A)

EP MCP Tool	Document Types Retrieved	Per-Document Analysis Trigger
get_adopted_texts	TA (adopted texts), legislative resolutions	Always — core legislative record
get_committee_documents	A* reports, opinions, working documents	When committee stage is scope
get_plenary_documents	Motions, amendments, agendas	When plenary activity is scope
get_procedures	Full procedure dossier	When tracking legislation
get_parliamentary_questions	QE (written), QO (oral)	When scrutiny activity is scope
get_external_documents	COM proposals, Council positions	When interinstitutional context needed
get_voting_records	Roll-call vote details	When vote analysis is scope
get_speeches	Plenary debate transcripts	When debate analysis is scope

⚠️ EP API date-filter note (v1.2.14+, Defect #5) — upstream contract active:

This is an upstream contract that is active on the pinned v1.3.12 server. Stage-B consumers may rely on the post-filter behavior described below.

Background: The EP Open Data Portal /meetings endpoint silently ignores its date-from / date-to query parameters. From european-parliament-mcp-server v1.2.14 onward (and on the pinned v1.3.12 server), the MCP server applies a server-side post-filter on dateFrom / dateTo before serialising get_plenary_sessions responses.

Consequences for Stage-B analysis on v1.2.14+ servers:

• total is the filtered count, not the raw upstream count. Do not treat total as representative of overall EP plenary activity when a date window is specified.
• Per-window session counts are reproducible. The EP-side regression is masked by the client-side filter, so repeated calls with the same dateFrom/dateTo will return consistent counts.
• No duplicate local filter exists in ep-mcp-client.ts. The upgraded upstream contract is the single source of truth for date-bounded session retrieval.

Downstream Consumption (Families A, C, D)

Per-document analyses feed into:

• Family A: synthesis-summary.md — aggregates per-document findings into executive narrative
• Family C: devils-advocate-analysis.md — sources ACH evidence from per-document fact extraction
• Family C: comparative-international.md — uses per-document EU-law mapping
• Family D: coalition-mathematics.md — derives vote counts from per-document roll-call analysis

---

🛠️ Per-Document Workflow

%%{init: {"theme":"dark","themeVariables":{"primaryColor":"#1565C0","primaryTextColor":"#ffffff"}}}%%
flowchart TD
    classDef input fill:#1565C0,stroke:#0D47A1,color:#FFFFFF
    classDef proc fill:#4CAF50,stroke:#1B5E20,color:#FFFFFF
    classDef gate fill:#FFC107,stroke:#F57F17,color:#3E2723
    classDef out fill:#7B1FA2,stroke:#4A148C,color:#FFFFFF

    A[Stage A — Data Retrieval<br/>EP MCP Tool calls]:::input
    B[Document Content Check<br/>Full text or metadata only?]:::gate
    C["Per-Document Analysis<br/>per-file-analysis/{doc-id}.md"]:::proc
    D[Document Index Update<br/>document-analysis-index.md]:::proc
    E[Family A/C/D Consumption<br/>Evidence flows upstream]:::out
    X["Skip — Metadata Only<br/>Record in manifest, no analysis"]:::gate

    A --> B
    B -->|Full text| C
    B -->|Metadata only| X
    C --> D
    D --> E
    X --> D

---

✅ Per-Document Completion Checklist

For each document analyzed:

• Full document content retrieved (verified via content hash)
• Document header block complete with Admiralty grade
• Executive summary ≤100 words, BLUF format
• Classification table has all 5 dimensions
• Clause-level analysis for legislative documents
• Actor mapping identifies ≥4 stakeholders
• Political group analysis covers all 8 groups
• Procedural Mermaid included for legislative docs
• Cross-reference section links related documents
• Forward indicators identify ≥3 monitoring points
• Source metadata complete with timestamp and hash

For the document analysis index:

• All per-document files linked
• Summary statistics table populated
• Cross-reference network Mermaid included
• Document list sorted by significance

---

📡 EP Open Data Portal Content-Availability Signal

Indexing Lag vs. Reliability Defect

The EP Open Data Portal publishes adopted-text identifiers (e.g., TA-10-2026-0104) to the /adopted-texts/feed endpoint before the full document content body is available. The gap between identifier publication and content availability is typically 5–15 days (documented in the EP Developer Corner).

The EP MCP Server (v1.2.13+) detects this sentinel and returns:

UPSTREAM_404: document indexed but content not yet available

This is a content-availability lag, not a retrieval reliability defect. The audit grader must not count these events as Defect #1 on every run.

Retry-Scheduling Sidecar

ep-mcp-client.ts getAdoptedTexts intercepts this sentinel and persists it to data/pending-documents.json with exponential back-off scheduling:

Attempt	Back-off delay	Notes
1st probe	24 h	Initial observation
2nd probe	48 h	First retry
3rd probe	72 h	Cap reached
4th+ probe	72 h	Capped
> 14 days	—	Escalated to wildcards-blackswans

Indexing-Lag Distribution (Empirical)

Audits tracking the delay between /adopted-texts/feed publication and content availability:

Observation Date	Affected IDs	Session	Lag Class
2026-04-24	TA-10-2026-0104, -0092, -0089, -0083, -0075, -0066, -0065, -0054, -0044, -0031, -0021; TA-10-2025-0345, -0336	propositions-run (13/13)	PENDING

As new runs resolve or escalate pending documents, append rows to the table above. This distribution informs:

• EP portal SLA expectations for Stage B deep-fetch planning
• Back-off calibration (if lag consistently < 24 h, reduce INITIAL_BACKOFF_MS)
• Escalation threshold (if lag consistently > 14 days for a batch, lower MAX_AGE_MS)

Workflow Integration Pattern

Stage B run N:
  1. client.escalateStalePendingDocuments()  // expire > 14-day docs
  2. dueIds = await client.getDueAdoptedTextsForReprobe()
  3. for each docId in dueIds:
       result = await client.getAdoptedTexts({ docId })
       if (realContentReturned(result)):
           await client.resolveAdoptedText(docId)
       // else: getAdoptedTexts already updated the back-off schedule
  4. console.log(await client.getPendingDocumentsSummary())

Audit Grader Guidance

When generating mcp-reliability-audit.md:

• Classify CONTENT_PENDING entries (from getFailedTools) separately from NOT_FOUND
• Report the count of pending documents and due-for-reprobe counts as availability-lag metrics
• Report escalated documents (status = ESCALATED) as potential data-completeness issues requiring wildcards-blackswans coverage — these are legitimate intelligence gaps
• Do not classify any CONTENT_PENDING event as a transport defect

---

Control	How this methodology satisfies it
ISO 27001 A.5.10 (Information classification)	Document classification table applies sensitivity ratings
ISO 27001 A.8.3 (Access control)	Source metadata provides audit trail for information provenance
NIST CSF ID.AM-3 (Data flows mapped)	Document cross-references map information flow
NIST CSF PR.DS-6 (Integrity checking)	Content hash enables integrity verification
CIS 3.4 (Data classification)	Political significance scoring classifies data
GDPR Art. 30 (Records of processing)	Source metadata provides processing record
NIS2 Art. 21(2)(d) (Supply chain security)	External document sourcing documented

---

📄 Document Control

Owner: CEO (Intelligence Program) · Reviewer: Chief Analyst · Review Cycle: Quarterly Next Review: 2026-07-31 · Related: ai-driven-analysis-guide.md, structural-metadata-methodology.md, artifact-catalog.md

---

Generated following EU Parliament Monitor Per-Document Methodology v1.0 — Family E Atomic Evidence Layer.