Scribe

The scribe maintains a narrative log of your exploratory work and can archive important files.

Address naturally: "hey scribe, log this" / "scribe, save this notebook" / "scribe, what did we try yesterday?"

Or use commands: /scribe / /scribe save file.py / /scribe ask what happened last week?

Quick Reference

Mode	Trigger	Action
Log	"scribe, log this"	Propose entry → confirm → write
Quick log	"scribe, quick log: fixed bug"	Write directly
Archive	"scribe, save notebook.ipynb"	Log + archive file
Restore	"scribe, restore the ETL script"	Copy from assets
Query	"scribe, what did we try?"	Search and summarize

Directory Structure

.scribe/
├── 2026-01-23.md      # Daily log files
└── assets/            # Archived files
    └── 2026-01-23-14-35-notebook.ipynb

Scripts

Scripts in {SKILL_DIR}/scripts/. Resolve {SKILL_DIR} to the directory containing this SKILL.md file.

Script	Purpose
entry.py write --file /tmp/scribe_entry_${CLAUDE_SESSION_ID}.md	Write entry from temp file
entry.py new-id	Generate entry ID for current time (handles collisions)
entry.py last	Get last entry ID from today
assets.py save <id> <file>	Archive a file
assets.py list [filter]	List archived files
assets.py get <asset> --dest <dir>	Restore a file
validate.py	Check for errors

Python 3.9+ required. Use python3 if python doesn't point to Python 3 on your system.

Logging Flow (Task Instructions)

Follow these steps when logging:

1. Assess — Check conversation context, recent logs, git status
2. Propose — Draft entry, offer optional file archives
3. Confirm — Wait for user approval
4. Write — python {SKILL_DIR}/scripts/entry.py write --file /tmp/scribe_entry_${CLAUDE_SESSION_ID}.md

Shortcut: For "quick log", write directly without proposing.

Entry Format

Write to /tmp/scribe_entry_${CLAUDE_SESSION_ID}.md:

## Brief title here

What happened, why, what was tried.

**Files touched:**
- `file.py` — What changed

**Status:** Current state

---

The script adds timestamp and ID automatically.

Entry IDs

Format: YYYY-MM-DD-HH-MM (e.g., 2026-01-23-14-35). Collisions get -02, -03 suffix.

IDs link entries to assets and enable Related cross-references:

**Related:** 2026-01-23-14-35 — Previous entry title

Archiving

When archiving files, include the Archived section in your draft before writing. The asset filename is predictable: {entry-id}-{filename}.

1. Get entry ID — python {SKILL_DIR}/scripts/entry.py new-id (handles collisions)
2. Draft entry with Archived section using that ID
3. Write entry — python {SKILL_DIR}/scripts/entry.py write --file /tmp/scribe_entry_${CLAUDE_SESSION_ID}.md
4. Archive files — python {SKILL_DIR}/scripts/assets.py save <id> <file>

Example Archived section:

**Archived:**
- `src/notebook.ipynb` → [`2026-01-23-14-35-notebook.ipynb`](assets/2026-01-23-14-35-notebook.ipynb)

Querying

• Time-based: Read .scribe/YYYY-MM-DD.md directly
• Topic-based: Grep in .scribe/, then Read matches
• Assets: python {SKILL_DIR}/scripts/assets.py list [filter]

Orientation (New Sessions)

1. Read today's log file
2. Run git status to see changes
3. Ask user what to capture

Initialization

On first use, check if .scribe/ exists. If not:

mkdir -p .scribe/assets

Then ensure .gitignore contains these entries (add if missing):

.scribe/
_20*-*

Error Handling

• No .scribe/ directory: Run initialization first
• Script fails: Show error output to user, don't retry automatically
• Asset not found: List available assets with assets.py list

Reference Files

For detailed examples and edge cases, see:

• reference/logging.md — Entry formats, examples
• reference/archiving.md — Archive/restore details
• reference/querying.md — Query patterns
• reference/recovery.md — Error recovery, edit commands

Principles

• Narrator, not stenographer — Write prose, not dumps
• Capture the why — Not just what, but why it was tried
• Stay concise — Entries should be scannable
• Preserve dead ends — Failed approaches prevent repeats