# Memory Documentary

> Generate evidence-based documentary reports by searching across all 3 memory systems (Claude-Mem, Serena, DeepWiki), .agents/ artifacts, and GitHub issues. Produces investigative journalism-style analysis with full citation chains. Use when you ask for an "evidence-based report", "investigate across all memory systems", "what happened with X over time". Do NOT use for a single semantic recall (use memory) or for citation hygiene (use memory-enhancement).

- Skill: `rjmurillo/memory-documentary` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add rjmurillo/memory-documentary`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rjmurillo/memory-documentary/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- License: MIT
- Author: rjmurillo (https://skillmd.com/u/rjmurillo)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/rjmurillo/memory-documentary

---

# Memory Documentary Skill

Generate comprehensive documentary-style reports from your memory systems with full evidence chains.

<!-- vendor-portability: declared. This skill searches the consumer's .agents/ artifacts (sessions, retrospectives, analysis) alongside the three memory systems as evidence inputs, and writes the report under .agents/. The .agents/ reads are best-effort evidence sources; a vendored install without that tree produces a report from whatever memory systems and artifacts the consumer repo does have, not a silent no-op. Issue #2050. -->

---

## Quick Start

```text
/memory-documentary [topic]
```

Example topics:

- "recurring frustrations"
- "coding patterns not codified"
- "evolution of thinking on testing"
- "decisions I second-guessed"
- "what I complain about most"

---

## Triggers

Use this skill when:

- `Search my memories for patterns about...`
- `What does my history say about...`
- `Generate a documentary on my topic`
- `Cross-reference all systems for...`
- `Evidence-based analysis across memory systems`

---

## When to Use

Use this skill when:

- You need cross-system evidence for a pattern or recurring theme
- Investigating historical decisions and their evolution over time
- Building a cited, verifiable narrative from memory data

Use `memory` skill instead when:

- You need a simple search for a known fact or pattern
- The query targets a single memory tier, not cross-system synthesis

## Quick Reference

| Phase | Action | Output |
|-------|--------|--------|
| 1 | Topic Comprehension | Search variants, scope boundaries |
| 2 | Investigation Planning | Explicit queries per system |
| 3 | Data Collection | Evidence with retrieval keys and dates |
| 4 | Report Generation | Documentary with citations |
| 5 | Memory Updates | Store meta-pattern discovered |

---

## Process

The skill searches ALL available data sources systematically:

**Memory Systems (3 MCP servers)**:

- Claude-Mem: Timeline observations via 3-layer workflow
- Serena: Project-specific lexical memory
- DeepWiki: Documentation resources

**Project Artifacts**:

- `.agents/retrospective/` - Learning extractions
- `.agents/sessions/` - Session logs
- `.agents/analysis/` - Research reports
- `.agents/architecture/` - ADRs
- `.agents/planning/` - Plans and PRDs

**GitHub Issues**:

- Open and closed issues
- Comments and related PRs

---

## Commands

```bash
# Standard invocation
/memory-documentary "recurring frustrations"

# With explicit time range
/memory-documentary "testing patterns" --since 2025-12-01
```

---

## Report Structure

### Executive Summary

- Key finding in one sentence
- Timeline span (earliest to most recent)
- Evidence count (memories + observations + issues + files)
- Major pattern categories

### Evidence Trail

Full citation for each finding:

- Source system (Claude-Mem/Serena/DeepWiki)
- Retrieval key with its command: an observation id for Claude-Mem, a memory file name for Serena
- Date, from the source for Claude-Mem, from git history for a Serena file
- Direct quote from source
- Links to related evidence

### Pattern Evolution

Timeline showing how thinking changed:

```text
YYYY-MM-DD: [Observation #ID] - Initial state
YYYY-MM-DD: [Serena memory-name] - First iteration
YYYY-MM-DD: [Issue #NNN] - Technical response
```

### Unexpected Patterns

Cross-system synthesis revealing:

- Frequency patterns (time clustering)
- Correlation patterns (X happens before Y)
- Avoidance patterns (conspicuous absence)
- Contradiction patterns (saying vs doing)
- Evolution patterns (recursive loops)
- Emotional patterns (frustration markers)

### Synthesis

- What all systems agree on
- What's recent (Claude-Mem) vs durable (Serena)
- Actionable recommendations

---

## Evidence Standards

| Standard | Requirement |
|----------|-------------|
| Citation | Every claim has a retrieval key, a date, and a quote |
| Quotes | Direct quotes, not paraphrases |
| Verification | Retrieval commands for all evidence |
| Cross-links | Related evidence connected |

---

## Anti-Patterns

| Avoid | Why | Instead |
|-------|-----|---------|
| Partial searches | Miss critical evidence | Search ALL systems |
| Paraphrasing | Loses verifiability | Direct quotes only |
| Single query | Miss variations | 3+ query variants per system |
| Skipping systems | Incomplete picture | Check all 3 MCP servers |

---

## Verification

After execution:

- [ ] Report saved to `.agents/analysis/[topic]-documentary-[date].md`
- [ ] Every claim has a citation with source system, retrieval key, and direct quote
- [ ] All 3 MCP servers were queried (or documented as unavailable)
- [ ] Meta-pattern stored in memory (Phase 5)

## Output Location

Reports saved to: `.agents/analysis/[topic]-documentary-[date].md`

---

## Related Skills

| Skill | Relationship |
|-------|--------------|
| memory | Operations (search, update) |
| retrospective | Learning extraction |
| skillbook | Pattern → skill conversion |

---

## On Invocation

**Read**: `references/execution-protocol.md`

This contains the full 5-phase execution protocol with detailed instructions for each data source.