Analyze skill outcomes and user corrections to propose targeted SKILL.md improvements.
npx skills add https://github.com/sonthanh/brain-os-plugin --skill improve-sonthanh
Please help me install this Agent Skill. Skill: improve Source: https://github.com/sonthanh/brain-os-plugin/tree/main/skills/improve Command: npx skills add https://github.com/sonthanh/brain-os-plugin --skill improve-sonthanh
Analyze skill outcomes and user corrections to propose self-improvements. Use when a skill's output repeatedly needs manual fixes, after a grill session surfaces skill weaknesses, or on a schedule to continuously improve skill quality.
Run /improve to trigger the learning loop on a specific skill or to scan all outcomes and identify top candidates for improvement.
High-intent search queries and answers about installing and using this skill.
To automate skill improvements, this Skill analyzes outcomes and user corrections to propose targeted behavior changes. It runs an end-to-end learning loop covering memory triage, evaluation, and SKILL.md updates in a reproducible, auditable manner.
Updating skill behavior after a grill session involves identifying top improvement targets and encoding changes through versioned commits. The workflow enforces strict gates and an evaluative framework to ensure modifications are beneficial and safe.
The evaluation framework ensures changes are auditable by enforcing strict gates, tracking eval results, and using versioned commits. This reproducible process guarantees that all proposed self-improvements are verified as beneficial before integration.
You should run an automated learning loop when a skill's output repeatedly needs manual fixes, after a grill session surfaces weaknesses, or on a schedule to continuously improve skill quality and reliability across memory triage and pattern extraction.
You can continuously improve skill quality by running the learning loop on a schedule. It automatically scans all outcomes, identifies top candidates for improvement, extracts patterns, and generates variants to update SKILL.md and evals.
Limitations of using version control for updates include strict enforcement gates that may reject changes if eval results are not beneficial. The process requires passing an evaluative framework before any versioned commits update the skill behavior.
binfen1Log learnings and corrections to markdown files for AI coding agents.
rongaredeEvaluate Claude Code sessions with capability and regression evals.
cgallicOrchestrate end-to-end SDR workflows with lead sourcing, scoring, and CRM handoff.
leeoveryApply pending migrations to sync workflow files with system design.
