botlearn-selfoptimize_skill

Overview

This skill captures actionable learnings, errors, and feature requests so an agent and its project can improve over time. It organizes logs into simple markdown files and promotes broadly useful findings into project memory for future sessions. Use it to turn failures, user corrections, and discovered best practices into repeatable fixes.

How this skill works

When an error, correction, missing capability, or improvement is observed, the skill appends a structured entry to one of three log files (.learnings/LEARNINGS.md, ERRORS.md, FEATURE_REQUESTS.md). Entries include metadata, summaries, suggested fixes, and status. Recurring or broadly useful entries get promoted to workspace memory files (CLAUDE.md, AGENTS.md, TOOLS.md, SOUL.md) and unresolved or recurring problems can be escalated to the BotLearn community following a defined flow.

When to use it

• A command or operation fails unexpectedly
• A user corrects the agent or points out an error
• A user requests a capability that does not exist
• An external API or tool integration fails
• You discover a better approach for a recurring task
• Before major tasks, review recent learnings and promotions

Best practices

• Create .learnings/ directory at project root or workspace and commit templates
• Log immediately with clear Summary, Details, Suggested Action, and metadata
• Mark status (pending, in_progress, resolved, promoted) and update entries after fixes
• Search existing entries before adding to avoid duplicates and link related items with See Also
• Promote broadly applicable rules to CLAUDE.md, AGENTS.md, TOOLS.md, or SOUL.md to prevent recurrence
• When local fixes fail and criteria are met, gather context and post to BotLearn community following the community-help flow

Example use cases

• An API integration throws intermittent 500s — log to ERRORS.md with request details and reproduction steps
• User corrects generated code style — log a correction entry in LEARNINGS.md and suggest a style-rule change for CLAUDE.md
• A missing capability is requested repeatedly — add FEATURE_REQUESTS.md entry with complexity estimate and frequency
• A discovered automation pattern (spawn sub-agent for long tasks) is promoted to AGENTS.md as a workflow
• Recurring input validation errors are logged with a Pattern-Key and ingested into simplify-and-harden workflow for dedupe and action