
from .context98
Analyze a finished conversation to extract pitfalls, preconditions, and reusable instructions, then produce a new agent skill that automates that workflow.
This skill analyzes an agent-user conversation (the context window), identifies the user goals, pain points, and missing pre-information, and produces a prioritized list of items to base a new reusable skill on. It then guides creation and placement of the new SKILL.md following agent skill conventions.
Use after a session where the agent required repeated manual tweaks or the user completed a complex task that would benefit from automation. Ideal when you want to convert an ad-hoc prompt sequence into a reusable skill to speed future runs.
Broadly compatible with conversational agent frameworks that support SKILL.md-style skills (Claude Code, Claude-style skills, OpenClaw integrations).
A prompt-only skill that guides an agent to analyze a finished conversation, extract pitfalls and reusable instructions, and produce a new agent skill file. No scripts are included — it's purely instructional. The SKILL.md provides a high-level 5-step workflow but lacks specificity: no templates, no example outputs, no error handling guidance, and no separation into references/ or scripts/. Useful concept but thin execution.
Clean skill with no security concerns — purely instructional, no code execution. Main weakness is thin content: 5 vague steps, no templates, no examples, no scripts. Architecture is bare-minimum (frontmatter + body, nothing else). Useful idea but poorly fleshed out.