Create Recursive Language Model skill epic #84
Conversation
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
Add comprehensive task epic for building RLM skill that applies MIT's Recursive Language Model strategies to AI agent codebase exploration. Epic includes 7 tasks covering research, design, implementation of fan-out search, SQLite indexing, and SudoLang skill specification.
Research findings include: - Agent Skills specification from agentskills.io - MIT RLM paper analysis (arXiv 2512.24601) - Implementation architecture for aidd framework - Complete SQLite database design with FTS5 Documents recursive language model strategies for deep codebase exploration through hierarchical task decomposition.
Replace single RLM epic with comprehensive phase-based approach: Phase 1: SQLite Foundation + Indexing + Fan-out Search - Database schema for documents, dependencies, productmanager - Frontmatter and dependency indexers with TDD - Fan-out search combining keyword, semantic, structured - CLI query and indexing tools Phase 2: Jiron DB API Skill - Generic skill for SudoLang → Jiron API generation - SQLite query bridge and middleware integration - Fan-out search exposed via Jiron endpoints - Includes ai/skills/jiron skill to teach agents Jiron protocol Phase 3: Frontmatter Index Jiron API - Apply Jiron DB API skill to frontmatter/codebase index - Self-documenting API with progressive discovery - Search and traversal endpoints with rich queries Phase 4: ProductManager Skill Conversion - Convert ai/rules/productmanager.mdc to full skill - Jiron API for story maps with SQLite as source of truth - Import existing YAML, export on demand - Database-backed workflows Phase 5: RLM Skill - Recursive exploration through hierarchical decomposition - Orchestrates all previous phases - User journey documented with pain points - Always-surfaced skill with agent self-direction Archived original RLM epic for reference.
Analyzed 2026-01-20-claude-code-skills branch to check for conflicts with our Phase 1-5 RLM + DB + Jiron epic plans. Key findings: - No blocking conflicts (.claude/ vs ai/ directories) - janhesters built YAML-based skills for Claude Code - Our approach adds SQLite + Jiron APIs + RLM on top - Different problem spaces that complement each other - Both can coexist and potentially converge later Recommendation: Proceed with all 5 phases as planned.
Created comprehensive overview document that provides: - Executive summary and big picture - Visual phase dependencies and implementation order - Concise summary of each phase with deliverables - Architecture decisions with rationale - Value proposition and quantified benefits - Success metrics for each phase - Implementation strategy and risk mitigation - Links to detailed phase epics and research docs - Relationship to janhesters' work Makes it easy to absorb the entire 5-phase epic structure at a glance before diving into detailed task lists.
Creates database-backed indexing system with FTS5 full-text search, dependency tracking, and ProductManager entities. Replaces YAML as source of truth for project metadata. Implements Phase 1 foundation for Recursive Language Model skill.
Auto-formatting from pre-commit hook.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Add comprehensive task epic for building RLM skill that applies MIT's
Recursive Language Model strategies to AI agent codebase exploration.
Epic includes 7 tasks covering research, design, implementation of
fan-out search, SQLite indexing, and SudoLang skill specification.
Note
Lays groundwork for the RLM skill with a SQLite indexing schema and planning docs, plus project setup updates.
ai/tools/create-schema.js(with CLI) and comprehensive tests to create SQLite schema:documents+ FTS5,dependencies, and ProductManager tables (personas,journeys,steps,pain_points,stories,functional_requirements), with indexes and triggersplan/rlm/(spec, architecture, database design, research summary)package.json(addsbetter-sqlite3, newtest:bun, adjusts unit test filters) and adds.gitattributesai/index.md,ai/rules/index.md) and addsai/rules/sudolang/index.mdandai/tools/index.mdWritten by Cursor Bugbot for commit 5bb8333. This will update automatically on new commits. Configure here.