skill-reinforcement

The skill-reinforcement skill functions as a self-improving meta-layer for autonomous agents, ensuring that every task performed contributes to the long-term intelligence of the system. Designed for AI agents operating within environments like 0 Finance, it bridges the gap between raw execution and continuous documentation. By automatically triggering after skill completion, it prevents knowledge loss, refines operational decision-making, and systematically updates local skill files to reflect the most efficient paths to success. This tool is essential for developers maintaining complex agent repositories who want to minimize token waste and maximize task reliability over time.

• Automatically analyzes success, failure, and performance metrics after every skill execution.

• Captures actionable learnings including shortcuts, anti-patterns, and environmental quirks.

• Systematically updates documentation files within .opencode/skill/*/SKILL.md to ensure persistent knowledge across sessions.

• Integrates with self-improve skills to manage decision trees, templates, and complex workflow logic.

• Provides a structured framework for documenting failure modes and effective fixes for recurring issues.

• Facilitates cross-pollination of knowledge across related skills like chrome-devtools-mcp, test-staging-branch, and financial workflows.

• Trigger this skill automatically whenever a task completes, a workflow fails, or a notable discovery is made.

• Follow the five-step reinforcement process: capture context, identify learnings, categorize findings, update the skill file, and validate the update.

• Use specific templates provided for token saving, failure modes, anti-patterns, and workflow improvements to ensure consistent formatting.

• Input expected includes skill name, task description, outcome (success/partial/failure), and performance metrics like token cost and latency.

• Constraints include maintaining non-redundant documentation and ensuring all updates are specific and actionable rather than vague observations.

• Regularly review the most frequently used skills to prioritize maintenance and identify when to transform repeatable tasks into dedicated tools or agents.