V1 Content Audit

Date: 2026-07-28

Verified Findings

Finding	File	Affected records	Evidence	Severity	Required remediation	Blocks v2.0.0
Template blocks were repeated	catalog/templates.json	49	Field-level check found 1 unique variables block, 1 unique required-inputs block, 1 unique expected-output block, 1 unique acceptance block, and 1 unique failure-mode block across 49 templates.	HIGH	Replace with task-specific templates.	Yes
Pattern blocks were repeated	catalog/patterns.json	26	Field-level check found 1 unique use_when, avoid_when, acceptance, failure-mode, and verification block across 26 patterns.	HIGH	Rewrite patterns around distinct mechanisms.	Yes
Prompt Contract variants were structurally duplicated	catalog/prompt_contracts.json	6	Hash check over dimensions and acceptance criteria found all six variants in one duplicate group.	HIGH	Add genuinely different copyable prompt structures.	Yes
Provider pages were too short	docs/providers/*.md	7 provider pages	Each provider page had 12 lines and only brief summary plus source links.	MEDIUM	Expand dated provider guidance with official sources and stale-risk notes.	Yes
JSON Schema files were not executed by a real validator	scripts/validate_catalog.py, schemas/*.json	All schemas	No jsonschema dependency or Draft202012Validator execution existed.	HIGH	Add real JSON Schema validation.	Yes
Schemas were permissive	schemas/*.json	7 schemas	Required content-quality fields were absent or shallow.	HIGH	Strengthen required fields and enums.	Yes
Makefile indentation was suspicious	Makefile	All targets	Targets were indented with spaces before target names; local make executable was unavailable, so execution could not be verified locally.	MEDIUM	Normalize Makefile and verify in CI.	Yes
Tests were insufficient	tests/*.py	3 tests	Tests did not detect field-level duplication, schema execution gaps, privacy, freshness, or semantic similarity.	HIGH	Add meaningful quality gates.	Yes
Count claims overstated unique substance	catalog/*.json	Patterns and templates	Numeric counts passed while repeated blocks made many records shallow.	HIGH	Prefer fewer substantive records over filler.	Yes
v1.0.0 should not be promoted broadly	GitHub release	v1.0.0	Release exists, but v1 quality defects above were verified.	MEDIUM	Keep v1 unchanged; prepare v2 through review PR.	No