Procedure Improvement
Overview
Systematically improve GOSM library procedures using schema-driven validation and tier-based progress tracking
Steps
Step 1: Load schema and procedures
- Read the procedure schema from schema_path
- Parse all procedure files from procedure_paths
- Create inventory of procedures to assess
Step 2: Validate each procedure
For each procedure:
- Check Bronze tier requirements (id, name, version, description, steps)
- Check Silver tier requirements (inputs, outputs, when_to_use, examples)
- Check Gold tier requirements (failure_modes, gosm_integration, verification)
- Record current tier and missing fields
Step 3: Calculate priority scores
For each procedure: priority = usage_weight × severity_weight / effort_estimate
Usage weights:
- core: 3, meta: 3, extracted: 2, domain-specific: 1
Severity weights:
- failing_schema: 3, missing_examples: 2, missing_advanced: 1
Effort estimate:
- single_field: 1, section_add: 2, major_rewrite: 5
Sort procedures by priority (highest first)
Step 4: Improve procedures
For each procedure in priority_queue (up to target count):
- Read current content
- Identify gaps from validation_results
- Generate content for missing fields:
- version: Add “1.0.0” if missing
- inputs: Infer from steps and description
- outputs: Infer from steps and purpose
- when_to_use: Derive from purpose
- examples: Create from procedure purpose
- failure_modes: Identify common problems
- gosm_integration: Determine workflow placement
- Merge new content with original (preserve all original content)
- Validate improved version
- Verify no degradation (tier >= original tier)
- Write improved file
Step 5: Generate improvement report
Create summary document with:
- Procedures analyzed
- Tier distribution (before and after)
- Improvements made per procedure
- Remaining gaps for future work
When to Use
- After adding new procedures to the library
- Periodically to improve library quality
- When procedures feel incomplete or inconsistent
- Before major GOSM version releases
Verification
- All improvements are schema-justified (not arbitrary)
- No procedure degrades (tier before <= tier after)
- Original content preserved in all cases
- Priority algorithm applied consistently