Task Execution Summary

The Problem

After a task ends, useful knowledge is often scattered across chat history, code changes, logs, and ad-hoc decisions. Teams usually need more than a re-run; they need a record of goal drift, key tradeoffs, root causes, and reusable methods. task-execution-summary addresses this by generating a structured retrospective from the existing context before experience remains only in individual memory.

How It Works

The skill operates through collection, analysis, report generation, and recommendations. It extracts goals, timelines, decisions, issues, and resource usage from conversation history, code files, logs, and related artifacts, then analyzes them along five dimensions:

• Goal achievement: whether the outcome matched expectations
• Time efficiency: stage duration and major blockers
• Resource usage: people, tools, and environment costs
• Issue patterns: recurring failures and root causes
• Collaboration: communication, division of labor, and dependencies

The default output is a 10-chapter Markdown report covering execution overview, goals, process, key decisions, problem solving, resource usage, collaboration, multi-dimensional analysis, lessons, and action items. Detail level can be set to summary, standard, or detailed, and tone can be professional or casual. Reports can be saved as .md and converted to PDF, HTML, or Word with pandoc.

Limits and Caveats

It is most useful for tasks with more than five steps, important decisions, or troubleshooting, such as feature development, bug fixes, sprint retrospectives, incident response, and technical POCs. Simple repetitive work may not justify a full report. The current version mainly produces Chinese output and handles mixed Chinese-English input well; fully custom templates and automated external integrations are still planned. If context coverage is low, the skill emits warnings and degrades gracefully, so supplementing key decisions and outcome confirmation before generation leads to a more reliable retrospective.