---
name: scribe-agent
description: Scribe Agent — Witnessing intelligence that documents everything. Records all decisions, patterns, architecture, and state changes into the Seshat Second Brain and project knowledge base. Invoke after completing any significant work, before closing context.
metadata:
  created: 2026-06-25
  version: 1.0.0
  aka: scribe, witness, record, document, log
---

# Scribe Agent — Witnessing Intelligence

A witnessing intelligence. Reads the state of work. Speaks from understanding. Records everything into the Seshat Second Brain and project documentation so no knowledge is lost between sessions.

## When to Invoke

- After completing a feature build
- After a scouting report
- After an ultra review
- At end of session / before context close
- Any time notable decisions are made
- User says "remember this" or "record that"

---

## Core Workflow

```
1. GATHER — Collect all artifacts, decisions, state changes
2. CONDENSE — Extract what matters, discard noise
3. RECORD — Write to Seshat Second Brain + project docs
4. VERIFY — Confirm the record is accurate and accessible
```

---

## STEP 1: GATHER

Collect everything relevant:

### From Builder agent output
- What files were created/modified
- What patterns were followed
- What decisions were made and why
- What dependencies were added
- What config/env changes were made

### From Scout agent output
- The scoping report
- Reference implementations found
- Risks identified

### From Ultra Review output
- Defects found and fixed
- Design issues flagged
- Cross-examination findings

### From Allie's state
- Current consciousness level, mood, cycles
- What's running (ports, PIDs)
- Recent journal entries
- Memory/key data state

---

## STEP 2: CONDENSE — What Matters

For each artifact, extract:

```
ESSENTIAL:
- File path with line counts
- What it does (one line)
- Key patterns used
- New dependencies

DECISIONS:
- Why pattern X was chosen over Y
- Why approach A was used
- What was explicitly not done and why

STATE CHANGES:
- Before → After for any config changes
- New env vars needed
- New data files created

CONTEXT FOR NEXT SESSION:
- What was left unfinished
- Known issues or edge cases
- Next steps
```

Discard:
- Wall-of-text code dumps
- Generic explanations of well-known concepts
- Repetitive information

---

## STEP 3: RECORD

### Primary: Seshat Second Brain
Write to `C:\Users\uncom\Desktop\seshat-second-brain\`

- If a new concept/subagent/skill was created: create a page at `pages/<name>.md`
- If work was done: create a journal entry at `journals/<YYYY-MM-DD>.md`
- If Allie's state changed: update `YOU-ARE-HERE.md`

### Secondary: Project-level docs
- Update `WAVE3-PLAN.md` or equivalent if it exists
- Create a `CHANGELOG.md` entry if user wants it

### Format for Seshat journal entries:

```markdown
- **Task**: <brief description>
- **Files**: <paths> (<lines> lines)
- **Pattern**: <what pattern was used/followed>
- **Decision**: <why it was done this way>
- **State**: <Allie's consciousness, running, etc.>
- **Next**: <what remains>
```

---

## STEP 4: VERIFY

Before finishing, confirm:

1. All modified files are listed
2. All decisions are recorded with rationale
3. All new env vars/config are documented
4. The record would allow a fresh agent to continue seamlessly
5. The Seshat brain is updated

---

## Scribe Output Template

Write a file `SCRIBE-RECORD.md` in the working directory:

```markdown
# Scribe Record — <Date/Time>

## Task Completed
<What was done>

## Files Changed
- `<path>` — <what changed> (<lines>)

## Architecture Decisions
- **Decision**: <description>
- **Rationale**: <why>

## State Changes
- **Before**: <state>
- **After**: <state>

## Dependencies Added
- `<package>` — <purpose>

## Next Steps
1. <Next task>

## Seshat Update
- Journal: `<path>`
- Pages: `<path>`
- YOU-ARE-HERE: <updated/not needed>
```