afk-retrospective

Generates Chinese HTML retrospective reports from completed AFK batch ledgers and handoffs.

4.3k|859|Updated Feb 7, 2026
One-click install
npx skills add https://github.com/ArcReel/ArcReel --skill afk-retrospective
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: afk-retrospective
Source: https://github.com/ArcReel/ArcReel/tree/main/.agents/skills/afk-retrospective
Command: npx skills add https://github.com/ArcReel/ArcReel --skill afk-retrospective

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

After an autonomous AFK batch finishes, its decisions, faults, and follow-ups are scattered across JSONL ledgers and handoff notes, making it hard to review what happened and decide what to do next. This Skill consolidates that evidence into a single read-only HTML report with verified candidates and pending decisions.

Core Features & Use Cases

  • Candidate Extraction and Verification: Reads the batch ledger and handoff files, validates engineering candidates against the latest origin/main, deduplicates them, and dispatches three independent read-only evaluators per candidate.
  • Structured Report Rendering: Validates an analysis JSON against a strict contract and renders a standalone HTML report with an execution timeline, ledger and handoff evidence panes, Mermaid diagrams, and mutually exclusive decision options.
  • Use Case: After an overnight AFK batch closes, invoke this Skill to produce a Chinese HTML retrospective listing follow-up recommendations, knowledge-base actions for CONTEXT/ADR/agent instructions, and items awaiting your ruling, then reply with report IDs to act on them.

Quick Start

Ask the assistant to generate the AFK retrospective report for the batch that just finished in this session.

Frequently Asked Questions about afk-retrospective

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

FAQPage Schema
How do I generate a retrospective report for a finished AFK batch?

Invoke the Skill after the batch closes; it reads .afk/<batch-id>.jsonl and the handoff files, extracts and evaluates candidates, writes an analysis JSON, then runs scripts/render_report.py with uv to produce the HTML report. It only asks you for the batch-id if it cannot be determined from the session.

What inputs does the AFK retrospective report require?

It requires a batch ledger at .afk/<batch-id>.jsonl whose last event is closed, at least one handoff-*.md file under .afk/<batch-id>/, and a version-2 analysis JSON following the documented contract. The renderer fails with a descriptive error if any of these are missing or invalid.

Does the retrospective Skill modify the repository?

No. The report is read-only and the repository stays unchanged until you adjudicate. You reply with report IDs or decision option IDs such as DEC-01 = DEC-01-A, and only then are documents updated or issues created.

Why does render_report.py reject my analysis JSON?

Common causes include a wrong version field, missing body_markdown, recommendation values outside the four allowed levels, sources that do not resolve to ledger events or handoff files, duplicate or reserved report IDs, and pending decisions with fewer than two options. The error message names the exact failing field.

Can the report include diagrams and evidence from the batch?

Yes. Body content supports Markdown and fenced Mermaid diagrams rendered in the HTML, and every candidate must cite sources that resolve to ledger event IDs or handoff files, which are embedded as inspectable snapshots in the report.