Marketplace for trusted AI agent skills

Skill Detail

Extract reviewable relationship timelines from chat exports with Afterglow

Analyze supplied chat exports into resumable JSON and Markdown relationship timelines, with block coverage and skipped-input diagnostics.

Data Extraction & TransformationMulti-Framework
Data Extraction & Transformation Multi-Framework Security Reviewed
⭐ 317 GitHub stars
COPY SKILL INSTRUCTIONS (OPTIONAL)
npx skills add agentskillexchange/skills --skill extract-relationship-timelines-from-chat-exports-with-afterglow Copy
Uses the third-party skills CLI, not an ASE-owned installer. Check your agent’s compatibility. This copies instructions; complete the upstream tool setup below separately.
View source Documentation
At a glance
Tools required
Git, Python 3.12.x, uv, Afterglow xuwen CLI, configured analysis block/final model endpoints, authorized chat export files
Install & setup
Install Python 3.12.x, uv and Git. Clone https://github.com/kldhsh123/Afterglow, enter Afterglow/backend, and run uv sync –extra dev. Follow the upstream configuration guide to configure the analysis block and final model endpoints/credentials in backend/.env; empty ANALYSIS_* settings fall back to the main chat model. Use local endpoints for both stages if offline processing is required. Supply authorized chat exports for one target person. From backend/, inspect the selection with uv run xuwen analyze chat.json –list-blocks –language en, then run uv run xuwen analyze chat.json –timeline-only –language en. No companion server or automatic messaging is needed for this CLI workflow.
Author
Afterglow Contributors
Publisher
Community
Last updated
Oct 20, 2026
Quick brief

Use this skill when a user supplies authorized chat exports and wants a reviewable relationship-event timeline as files. The agent selects the explicit input files and date range, inspects deterministic analysis blocks, runs Afterglow’s offline timeline analysis, and returns the timeline with a coverage and failure summary. Invoke it for batch archive analysis or resumable report generation rather than an ordinary companion-chat turn.

How it works

What this skill actually does

Prerequisites: Git, Python 3.12.x, uv, the Afterglow source checkout, and configured block/final analysis model endpoints. Follow the upstream configuration guide for identities and environment settings. All selected exports must concern the same target person. Use supported exports documented upstream; the examples below use JSON/JSONL. Obtain permission to process the shared chats and confirm which model providers receive them.

git clone https://github.com/kldhsh123/Afterglow
cd Afterglow/backend
uv sync --extra dev
# Configure backend/.env and analysis providers before processing exports.
uv run xuwen analyze chat.json --list-blocks --language en
uv run xuwen analyze chat.json --timeline-only --language en
# Multiple files for the same target, optionally limited by month:
uv run xuwen analyze source-a.json source-b.jsonl --timeline-only --since 2025-01 --language en

Read results from ANALYSIS_DATA_DIR, normally backend/.data/analysis/. Return timeline.json and timeline.md together with manifest.json coverage/cache statistics and any failures/*.json diagnostics. Inspect reported events against the supplied source history; distinguish model-inferred stages from explicit dated evidence. Refused or invalid blocks may be skipped, so disclose missing coverage and never equate a completed run with complete extraction. Repeating a run can reuse successful block caches; keep the output language consistent.

The boundary is supplied exports → explicit timeline-only CLI analysis → reviewable report artifacts. Personality, Life and experimental reports, companion deployment, chat generation, proactive messaging and runtime prompt injection are outside this skill. Analysis reads original exports directly; a vector-store import is not a prerequisite. Cloud analysis can transmit selected chat blocks; configure both analysis stages locally if the user requires offline inference.

Framework: Multi-Framework. This is a standalone Python CLI with configurable model services and file outputs, callable from any agent runtime with shell access; no particular coding harness, model vendor or MCP server is required. Metadata and documented invocation were verified against upstream sources; installation, runtime behavior and security were not tested.