Capture and categorize user adjustment reasons to refine agent memory.
npx skills add https://github.com/coma-toast/inquisitive --skill inquisitive
Please help me install this Agent Skill. Skill: inquisitive Source: https://github.com/coma-toast/inquisitive/tree/main Command: npx skills add https://github.com/coma-toast/inquisitive --skill inquisitive
Inquisitive reduces first-try errors by learning why the agent’s proposed output didn’t match what the user actually wanted, based on real user adjustments.
When the agent proposes a change and you modify it afterward, say what you wanted instead and why, so inquisitive can learn your preference from that adjustment.
High-intent search queries and answers about installing and using this skill.
Agent memory for preference learning stores user adjustment reasons across repo, org, and user scopes using categorized entries, ensuring repeated preferences become durable context to improve first-try accuracy on subsequent tasks.
To set up agent memory for capturing plan modifications, configure your storage backends to persist categorized user feedback, enabling the agent to ask targeted why questions and save responses for later context loading.
Yes, this agent memory system supports scoped storage backends for different codebases by escalating stored preferences across repo, org, and user scopes to ensure correct context loading based on the active environment.
The best way to classify user feedback for meta-learning is sorting adjustment reasons into 12 distinct categories, allowing the agent to consolidate entries into evolving summaries and generate refinement sub-skills when strong patterns emerge.
You should use category-based context loading for refinement requests when an agent's proposed output repeatedly mismatches user expectations, requiring durable memory to apply previously captured adjustment reasons to new file edits.
dev-giosClassifies user dismissals of code review findings and persists feedback to Engram or OpenSpec.
breddinIntegrate ReasoningBank with AgentDB for trajectory tracking and memory distillation.
delorenjGuide Letta development for stateful AI agents with long-term memory.
alexandrbasisGuide users step-by-step through learning tasks with adaptive explanations.
