skills.rest
spikelab
spikelabCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

backfill

Reconstruct learnings and diary entries from existing Claude Code session transcripts.

Updated May 3, 2026
Search Tags:#claude code#memory#backfill#jsonl#diary#learnings extraction#session transcripts
One-click install
npx skills add https://github.com/spikelab/multiplai-cc-mktplace --skill backfill-spikelab
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: backfill
Source: https://github.com/spikelab/multiplai-cc-mktplace/tree/main/plugins/multiplai-context/skills/backfill
Command: npx skills add https://github.com/spikelab/multiplai-cc-mktplace --skill backfill-spikelab

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When you install the multiplai-context plugin, your memory files start empty even though you have months of Claude Code session history. This Skill scans your existing session transcripts and rebuilds learnings, diary entries, and now/ summaries so your persistent memory has historical depth from day one. ## Core Features & Use Cases - Transcript Scanning: Reads $CLAUDE_CONFIG_DIR/projects/**/*.jsonl session transcripts and distills them through the same diary-first extraction pipeline used by live sessions. - Flexible Time Windows: Defaults to the last 7 days, with --days N, --since DATE, --all, and --projects flags to control scope and cost. - Dry-Run with Cost Estimate: Shows session count, estimated tokens, and a privacy notice before any processing begins. - Idempotent Execution: Skips sessions already present in learnings/ and diary/, so reruns only fill gaps. - Use Case: After installing the plugin, run a dry-run over the last 30 days, confirm the token estimate, then launch the background backfill and consolidate results with /multiplai-context:dream and /multiplai-context:dream-remember. ## Quick Start Ask the assistant to backfill learnings from your Claude Code session history for the last 30 days, starting with a dry-run to review the estimated cost.

Frequently Asked Questions about backfill

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I backfill memory from past Claude Code sessions?▼

Run the backfill script with uv against your session transcripts, starting with --dry-run to see session count and token estimates. After confirming, execute it in the background and then run /multiplai-context:dream and /multiplai-context:dream-remember to consolidate results.

How do I limit backfill to specific projects or dates?▼

Use --projects with comma-separated project slugs to scope by repository, and --days N, --since DATE, or --all to control the time window. The default window is the last 7 days.

Does backfill reprocess sessions that were already extracted?▼

No, backfill is idempotent. Sessions already present in the learnings/ and diary/ directories are skipped, so rerunning it only fills gaps for unprocessed sessions.

Why does my backfill process never report completion?▼

Completion must be detected by the sentinel line 'Backfill complete' in the captured output, never by process-liveness checks. In this environment finished scripts become defunct zombies that match ps checks forever, so polling loops never terminate.

How much does a Claude Code transcript backfill cost?▼

Cost is roughly one LLM call per chunk with about a 32k token budget per chunk. The dry-run mode displays the session count and estimated tokens before you commit to the full run.

Related Skills

Explore More Software Engineering Skills→
dundasdundas

Transcript Query

Query Claude Code session transcripts in JSONL files for project history.

Community
Intermediate
bitranoxbitranox
1

meta-dream-nap

Consolidates session learnings into the current project's memory chain after context compaction.

Community
Advanced
polycode-publicpolycode-public

news-feed-quality

Iterates news card generation cycles to close gaps against a hand-written target card.

Official
Advanced
SethGammonSethGammon
820

cost

Extract token usage, session costs, and campaign spend from Claude Code session data.

Community
Intermediate