feat: Add self-improving skill system - #20
Open
YrFnS wants to merge 1 commit into
Open
Conversation
Implements Hermes-Agent's closed learning loop pattern for automatic skill creation and improvement from successful task patterns. ## New Tools - skill_manage: Create, edit, patch, delete, write-file, remove-file - skill_view: View skill full content with metadata ## New Modules - src/skills/utils.ts: Validation helpers (validateSkillName, validateSkillContent, findSkill, etc.) - src/skills/review.ts: SkillTracker + triggerSkillReview for automatic review suggestions - src/capabilities/skills/skill-manage.ts: Full skill_manage tool implementation with atomic writes - src/capabilities/skills/skill-view.ts: Skill viewing tool ## Features - Auto-suggest skills after 10 messages without skill actions - Survey existing skills before creating new (prefer patching) - Token-conscious async review generation (max 800 tokens) - Atomic writes with temp files (crash-safe) - Path validation and local skill detection - Automatic discovery refresh after skill changes ## Following Mercury's Contributing Guidelines - Agentic expertise: loop-aware, permission-first, token-conscious - Code quality: TypeScript, JSDoc comments, async/await, tests - No new dependencies added - Graceful degradation on validation/provider errors - Channel-agnostic (works on CLI and Telegram) ## Reference Based on Hermes-Agent's skill_manager_tool.py implementation (NousResearch/hermes-agent) ## Tests Added 10 tests covering: - Skill name/content validation - Skill finding and discovery - SkillTracker iteration tracking - Skill name/description suggestion
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Implements Hermes-Agent's closed learning loop pattern for automatic skill creation and improvement from successful task patterns.
New Tools
New Modules
Features
Following Mercury's Contributing Guidelines
Reference
Based on Hermes-Agent's skill_manager_tool.py implementation (NousResearch/hermes-agent)
Tests
Added 10 tests covering: