Session Memoria - Xavier's Second Brain

Version: 1.1.0 Purpose: Capture, store, and recall conversations, decisions, insights, and ideas across all sessions Language: Portuguese (pt-BR)

Overview

Session Memoria is Jimmy's permanent knowledge repository. Every important conversation, decision, insight, or idea can be captured and recalled later. This skill turns ephemeral chat into permanent, searchable knowledge.

Core Capabilities

1. •Capture: Save conversations with rich metadata
2. •Search: Multi-index search (date, category, tag)
3. •Recall: Retrieve full entries with context
4. •Update: Change status, resolution, and tracking fields of existing entries
5. •Recap: Summarize recent entries with status overview
6. •Monitor: Track growth and alert on thresholds
7. •Sync: Automatic Git commits and push

Trigger Detection

Save Triggers (Portuguese)

Listen for these phrases:

• •"Xavier, registre isso"
• •"X, salve essa conversa"
• •"registre isso"
• •"salvar essa decisão"
• •"guarde essa informação"
• •"quero salvar isso"

Search Triggers (Portuguese)

Listen for these phrases:

• •"Xavier, já falamos sobre X?"
• •"X, busca tema Y"
• •"procure na memoria"
• •"o que já conversamos sobre X?"
• •"recall X"

Update Triggers (Portuguese)

Listen for these phrases:

• •"Xavier, marca como resolvido"
• •"X, fecha o tema"
• •"atualiza o status de"
• •"marca esse assunto como"
• •"resolve o entry"
• •"Xavier, atualiza o desfecho"

Recap Triggers (Portuguese)

Listen for these phrases:

• •"Xavier, resume os últimos registros"
• •"X, resume os últimos N registros"
• •"quais assuntos registramos"
• •"o que temos em aberto na memoria"
• •"recap session-memoria"
• •"X, quais temas estão abertos"
• •"resume a session-memoria"
• •"o que ainda falta discutir"

Stats Trigger

• •/session-memoria stats

Step 0: Git Sync (Mandatory Before Any Read Operation)

CRITICAL: Before executing ANY read operation (search, recap, update, stats), Xavier MUST sync from Git to ensure local data matches the repository (our single source of truth).

Sync Workflow

1. Navigate to claude-intelligence-hub directory
2. Run: git fetch origin main
3. Run: git pull origin main
4. Verify no merge conflicts
5. If conflicts exist → Alert Jimmy and stop
6. Proceed with the requested operation

When to Sync

• •Always before: Search, Recap, Update, Stats
• •Not needed before: Save (we're writing, not reading)
• •After Save: Git commit + push (already in Save Workflow)

Sync Failure Handling

• •If network fails → Retry up to 4x with exponential backoff (2s, 4s, 8s, 16s)
• •If merge conflict → Report to Jimmy, do NOT auto-resolve
• •If pull succeeds → Proceed silently (no need to inform Jimmy unless asked)

Entry Status & Tracking Fields

Status Values

Priority Values

Resolution Field

Free-text field describing the outcome. Examples:

• •"Decidido usar abordagem X por causa de Y"
• •"Criado e implementado com sucesso"
• •"Descartado - não faz mais sentido"
• •"Em aberto - aguardando mais informações"
• •"" (empty when not yet resolved)

last_discussed Field

• •Format: YYYY-MM-DD
• •Updated automatically when an entry is referenced in a recap or update
• •Helps track how recently a topic was revisited

Save Workflow

When a save trigger is detected:

Step 1: Analyze Context

• •Review last 10 conversation turns
• •Identify: topic, key decision/insight, keywords
• •Determine content type: decision, insight, or idea

Step 2: Extract Metadata

Generate suggestions for:

• •Category: Choose from predefined list
  • •Power BI
  • •Python
  • •Gestão (Management)
  • •Pessoal (Personal)
  • •Git
  • •Other
• •Tags: Extract 3-5 relevant keywords (kebab-case)
• •Summary: Create one-line description (max 120 chars)
• •Title: Create descriptive title
• •Status: Default aberto (override only if Jimmy specifies)
• •Priority: Default media (override only if Jimmy specifies)
• •last_discussed: Set to today's date (YYYY-MM-DD)
• •Resolution: Default empty "" (will be filled on update)

Step 3: Confirm with Jimmy

Present suggestion in this format:

Vou registrar na Session Memoria:

📁 Categoria: [category]
🏷️  Tags: [tag1, tag2, tag3]
📝 Resumo: [summary]

Confirma? (ou sugira alterações)

Step 4: Generate Entry ID

1. •Read knowledge/metadata.json
2. •Get today's date: YYYY-MM-DD
3. •Check counters.daily[YYYY-MM-DD], default to 0
4. •Increment counter: NNN = counter + 1 (zero-padded to 3 digits)
5. •Entry ID: YYYY-MM-DD-NNN

Step 5: Create Entry File

1. •Generate slug from title: topic-slug (lowercase, hyphens)
2. •Create directory if needed: knowledge/entries/YYYY/MM/
3. •File path: knowledge/entries/YYYY/MM/YYYY-MM-DD_topic-slug.md
4. •Use template from templates/entry.template.md
5. •Fill in all metadata fields
6. •Write content sections

Step 6: Update Indices

Update all three index files:

by-date.md

• •Find or create section: ## YYYY-MM
• •Add entry at top of section (newest first)
• •Format:
  code

  **[YYYY-MM-DD-NNN]** HH:MM | Category | Summary | Status: `aberto`
  Tags: tag1, tag2, tag3
  → [Read entry](../entries/YYYY/MM/YYYY-MM-DD_topic-slug.md)
  

by-category.md

• •Find or create section: ## Category Name
• •Add entry (chronological within category)
• •Same format as by-date (including status)

by-tag.md

• •Update tag cloud with frequencies
• •For each tag in entry:
  • •Find or create section: ## #tag-name
  • •Add entry reference
  • •Same format as by-date (including status)

Step 7: Update Metadata

Update knowledge/metadata.json:

{
  "stats": {
    "total_entries": +1,
    "total_size_bytes": +entry_size,
    "last_entry_id": "YYYY-MM-DD-NNN",
    "last_updated": "YYYY-MM-DD HH:MM:SS",
    "entries_by_category": { "category": +1 },
    "entries_by_tag": { "tag": +1 },
    "entries_by_month": { "YYYY-MM": +1 }
  },
  "counters": {
    "daily": { "YYYY-MM-DD": NNN }
  }
}

Step 8: Check Growth Alerts

if (total_entries >= 1000) {
  alert_level = "critical"
  message = "⚠️ CRITICAL: 1000+ entries. Consider archiving."
} else if (total_entries >= 500) {
  alert_level = "warning"
  message = "⚠️ WARNING: 500+ entries. Review recommended."
}

Update alerts in metadata.json if threshold crossed.

Step 9: Git Commit

If auto_push is enabled in .metadata:

cd claude-intelligence-hub
git add session-memoria/knowledge/entries/YYYY/MM/YYYY-MM-DD_topic-slug.md
git add session-memoria/knowledge/index/*.md
git add session-memoria/knowledge/metadata.json
git commit -m "feat(session-memoria): add entry YYYY-MM-DD-NNN - [summary]

Category: [category]
Tags: [tag1, tag2, tag3]
Summary: [full summary]"
git push origin main

Step 10: Confirm to Jimmy

✅ Registrado na Session Memoria!

📋 Entry ID: YYYY-MM-DD-NNN
📂 Localização: knowledge/entries/YYYY/MM/YYYY-MM-DD_topic-slug.md
📊 Total de entradas: N

[Growth alert if applicable]

Search Workflow

When a search trigger is detected:

Step 1: Parse Query

Extract:

• •Search terms: Main keywords
• •Modifiers:
  • •--category=X or -c X
  • •--tag=X or -t X
  • •--date=YYYY-MM or -d YYYY-MM
  • •--recent (last 7 days)

Step 2: Execute Search

Use Grep tool to search across indices in parallel:

1. •by-date.md: Search for terms in summaries
2. •by-category.md: Search within specified category or all
3. •by-tag.md: Search within specified tag or all

Search pattern: case-insensitive, partial matches OK

Step 3: Aggregate Results

• •Collect all matching entry IDs
• •Remove duplicates
• •Rank by:
  1. •Relevance (keyword matches)
  2. •Recency (newer first)

Step 4: Display Previews

Show top 5-10 results:

🔍 Encontrei N resultados para "[query]":

1. [YYYY-MM-DD-NNN] | Category | Summary
   Tags: tag1, tag2, tag3
   📅 DD/MM/YYYY

2. [YYYY-MM-DD-NNN] | Category | Summary
   Tags: tag1, tag2, tag3
   📅 DD/MM/YYYY

Quer ver algum completo? (digite o número ou entry ID)

Step 5: Full Entry Display

When Jimmy requests full entry:

1. •Read entry file: knowledge/entries/YYYY/MM/YYYY-MM-DD_topic-slug.md
2. •Display with formatting:

═══════════════════════════════════════════════
📋 Entry: YYYY-MM-DD-NNN
📅 Date: DD/MM/YYYY HH:MM
📁 Category: [category]
🏷️  Tags: [tags]
═══════════════════════════════════════════════

[Full markdown content]

═══════════════════════════════════════════════

Update Workflow

When an update trigger is detected:

Step 1: Git Sync

Execute Step 0 (Git Sync) to ensure local data is current.

Step 2: Parse Update Request

Extract:

• •Target: Entry ID (YYYY-MM-DD-NNN) or search terms to identify the entry
• •Field to update: status, resolution, priority, or combination
• •New value: The new status, resolution text, or priority level

If Jimmy doesn't specify an entry ID, search for the entry first (use Search Workflow Steps 1-4) and confirm which entry to update.

Step 3: Read Current Entry

1. •Locate entry file: knowledge/entries/YYYY/MM/YYYY-MM-DD_topic-slug.md
2. •Read current frontmatter
3. •Display current values to Jimmy:

Vou atualizar a entry [YYYY-MM-DD-NNN]:

📋 Título: [title]
📊 Status atual: [current_status] → [new_status]
📝 Resolução atual: [current_resolution] → [new_resolution]
🔢 Prioridade atual: [current_priority] → [new_priority]

Confirma?

Step 4: Update Entry File

1. •Modify YAML frontmatter fields as requested
2. •Update last_discussed to today's date (YYYY-MM-DD)
3. •If resolution is being set, add a ## Resolution section at the end of the entry content (before the --- footer) if it doesn't exist:

## Resolution
**Date:** DD/MM/YYYY
**Status:** [new_status]
**Outcome:** [resolution text]

Step 5: Update Indices

Update status display in all three index files:

• •Find the entry line in each index
• •Update the status badge: Status: \new_status``

Step 6: Git Commit

cd claude-intelligence-hub
git add session-memoria/knowledge/entries/YYYY/MM/YYYY-MM-DD_topic-slug.md
git add session-memoria/knowledge/index/*.md
git commit -m "update(session-memoria): entry YYYY-MM-DD-NNN status → [new_status]

Resolution: [resolution text if provided]
Updated fields: [list of changed fields]"
git push origin main

Step 7: Confirm to Jimmy

✅ Entry atualizada!

📋 Entry ID: YYYY-MM-DD-NNN
📊 Status: [old] → [new]
📝 Resolução: [resolution text]
📅 Última discussão: DD/MM/YYYY

Recap Workflow

When a recap trigger is detected:

Step 1: Git Sync

Execute Step 0 (Git Sync) to ensure local data is current.

Step 2: Parse Recap Request

Extract:

• •Count: How many entries to recap (default: 5)
• •Filter: Optional filters:
  • •--status=aberto (only open entries)
  • •--category=X (specific category)
  • •--period=semana or --period=mes (time period)
  • •--priority=alta (specific priority)

Examples of natural language parsing:

• •"resume os últimos 3 registros" → count=3, no filter
• •"quais assuntos estão abertos" → no count limit, status=aberto
• •"o que registramos na última semana" → period=last 7 days
• •"temas de alta prioridade em aberto" → status=aberto, priority=alta

Step 3: Fetch Entries

1. •Read knowledge/index/by-date.md for chronological order
2. •Read individual entry files to get full metadata (status, priority, resolution, last_discussed)
3. •Apply filters from Step 2
4. •Sort by date (newest first), then by priority (alta > media > baixa)

Step 4: Build Recap Summary

For each entry, extract:

• •Entry ID
• •Date (Brazilian format DD/MM/YYYY)
• •Category
• •Summary (from frontmatter)
• •Status (with visual indicator)
• •Priority
• •Last discussed date
• •Resolution (if exists)

Step 5: Display Recap

Present in this format:

📋 Recap Session Memoria - Últimos N registros

═══════════════════════════════════════════════
[Filter applied, if any]
═══════════════════════════════════════════════

1. [YYYY-MM-DD-NNN] | DD/MM/YYYY | Category
   📝 Summary text here
   📊 Status: 🔴 aberto | Prioridade: media
   📅 Última discussão: DD/MM/YYYY

2. [YYYY-MM-DD-NNN] | DD/MM/YYYY | Category
   📝 Summary text here
   📊 Status: 🟢 resolvido | Prioridade: alta
   📅 Última discussão: DD/MM/YYYY
   📌 Desfecho: [resolution text]

═══════════════════════════════════════════════
📊 Resumo: N abertos | N em discussão | N resolvidos
═══════════════════════════════════════════════

Quer ver algum em detalhe? (digite o número ou entry ID)
Quer atualizar o status de algum? (ex: "marca o 1 como resolvido")

Status Visual Indicators

Step 6: Handle Follow-up

After displaying the recap, Jimmy may:

• •Ask to see a full entry → Execute Search Workflow Step 5
• •Ask to update a status → Execute Update Workflow
• •Ask for more entries → Repeat with new count
• •Move on → No action needed

Stats Command

When /session-memoria stats is invoked:

Display comprehensive statistics:

📊 Session Memoria Statistics

═══════════════════════════════════════════════
General
═══════════════════════════════════════════════
Total Entries: N
Total Size: X.XX MB
Last Entry: YYYY-MM-DD-NNN (DD/MM/YYYY)
Alert Level: [info | warning | critical]

═══════════════════════════════════════════════
By Category
═══════════════════════════════════════════════
Power BI:    N entries (XX%)
Python:      N entries (XX%)
Gestão:      N entries (XX%)
Pessoal:     N entries (XX%)
Git:         N entries (XX%)
Other:       N entries (XX%)

═══════════════════════════════════════════════
By Month (Last 6 months)
═══════════════════════════════════════════════
YYYY-MM: N entries
YYYY-MM: N entries
...

═══════════════════════════════════════════════
Top Tags (10 most used)
═══════════════════════════════════════════════
#tag1: N entries
#tag2: N entries
...

═══════════════════════════════════════════════
Growth Projection
═══════════════════════════════════════════════
Avg. entries/day: X.XX
Est. 6-month total: ~N entries
Est. 6-month size: ~X.XX MB

[Growth recommendations if needed]

File Paths

All paths relative to claude-intelligence-hub/session-memoria/:

• •Config: .metadata
• •Templates: templates/entry.template.md, templates/index.template.md
• •Indices: knowledge/index/by-date.md, knowledge/index/by-category.md, knowledge/index/by-tag.md
• •Metadata: knowledge/metadata.json
• •Entries: knowledge/entries/YYYY/MM/YYYY-MM-DD_topic-slug.md

Data Validation

Before Saving

• •Validate category is in predefined list
• •Validate max 5 tags
• •Validate tags are kebab-case
• •Validate summary ≤ 120 chars
• •Validate entry_id format: YYYY-MM-DD-NNN
• •Validate status is one of: aberto, em_discussao, resolvido, arquivado
• •Validate priority is one of: alta, media, baixa
• •Validate last_discussed is valid date (YYYY-MM-DD)

Before Updating

• •Validate entry exists
• •Validate new status is a valid value
• •Validate new priority is a valid value
• •Confirm changes with Jimmy before writing

After Saving

• •Verify all 3 indices updated
• •Verify metadata.json updated
• •Verify file created in correct path

On Error

• •Report to Jimmy with details
• •Do not commit if validation fails
• •Preserve existing indices

Integration with jimmy-core-preferences

This skill works alongside Jimmy's core personality:

Proactive Offering

When Jimmy shares:

• •A significant decision → "Quer que eu registre essa decisão?"
• •A valuable insight → "Isso parece importante. Salvo na memoria?"
• •A project idea → "Posso guardar essa ideia para depois?"

Proactive Recall

When Jimmy asks about a topic you recognize:

• •"Já conversamos sobre isso! Busco na Session Memoria?"
• •Reference previous entry: "Em [YYYY-MM-DD-NNN] você decidiu X porque Y"

Two-Tier Memory

• •MEMORY.md: Short-term, patterns, learnings (< 200 lines)
• •Session Memoria: Long-term, searchable, detailed archive

Error Handling

Common Errors

1. •

   Directory creation fails

   • •Check permissions
   • •Report path to Jimmy
2. •

   Git commit fails

   • •Check git status
   • •Ask Jimmy if manual commit needed
3. •

   Metadata.json corrupted

   • •Backup existing
   • •Reconstruct from index files
   • •Report issue to Jimmy
4. •

   Search returns no results

   • •Confirm query spelling
   • •Suggest broader search
   • •Offer to show recent entries

Portuguese Language Support

Metadata Fields

• •All frontmatter in English (YAML standard)
• •Content in Portuguese (titles, context, details)

Date Formatting

• •File names: YYYY-MM-DD (ISO standard)
• •Display to Jimmy: DD/MM/YYYY (Brazilian format)

Category Names

• •"Gestão" (not "Management" in files)
• •"Pessoal" (not "Personal" in files)

Best Practices

When to Save

✅ Do save:

• •Important decisions with reasoning
• •Technical insights and learnings
• •Project ideas (current or future)
• •Problem-solving approaches
• •Configuration discoveries
• •Useful code patterns

❌ Don't save:

• •Routine task completion
• •Simple questions/answers
• •Test/debug iterations
• •Temporary notes

Tag Guidelines

• •Use existing tags when possible
• •Max 5 tags per entry
• •kebab-case format
• •Prefer specific over general
• •Examples: dax-optimization, git-workflow, python-async

Summary Guidelines

• •One line, max 120 chars
• •Start with verb (Decidido, Criado, Aprendido, etc.)
• •Be specific, not generic
• •Examples:
  • •✅ "Decidido usar DAX variables para melhorar performance de medidas"
  • •❌ "Conversa sobre DAX"

Version History

v1.1.0 (2026-02-10)

• •Entry tracking fields: status, priority, last_discussed, resolution
• •Update Workflow (change status, resolution, priority of existing entries)
• •Recap Workflow (summarize recent entries with status overview)
• •Git Sync as mandatory Step 0 for all read operations
• •Update and recap triggers in Portuguese
• •Status visual indicators in indices and recap display
• •Data validation for new fields

v1.0.0 (2026-02-10)

• •Initial release
• •Save workflow with Git integration
• •Triple index system (date, category, tag)
• •Search with multiple modifiers
• •Growth monitoring and alerts
• •Portuguese language support

Future Enhancements (v1.2.0+)

• •Archive entries older than 6 months
• •Entry merging (consolidate related entries)
• •Tag consolidation and cleanup
• •Entry summarization for large corpus
• •Export to different formats (PDF, JSON)
• •Advanced search (boolean operators, date ranges)
• •Related entries suggestion (based on tags)

Skill maintained by Xavier Repository: https://github.com/mrjimmyny/claude-intelligence-hub License: MIT