[Feat]: CI doc-drift guards — skill catalog and examples index parity

[rosspeili]

Feature Description

Registry skills frequently land with correct bundles but stale or missing index rows in contributor-facing docs. Today this is caught only in manual review (e.g. missing agent_loops.md row for defi/evm_tx_handler, incomplete CONTRIBUTING category examples).

Add tests/test_registry_docs.py (framework layer, runs in pytest tests/) with automated parity checks:

1. Catalog parity — count of skills/**/manifest.yaml equals skill rows in docs/skills/README.md (parse table rows with backtick skill IDs; exclude section headers).
2. Catalog page exists — each manifest has docs/skills/<skill_name>.md (derive skill_name from path).
3. Examples index parity — every examples/*.py (except *_common.py if grandfathered) appears in examples/README.md Runnable Scripts table; flag orphaned index rows pointing to missing files.
4. Agent loops matrix (soft check) — every registry skill ID appears at least once in docs/usage/agent_loops.md table (may be (catalog page) or script name).

On first add, fix any failures or document a short grandfather list with TODO + issue links. Relates to #104 (provider sections on catalog pages) but does not duplicate it.

Rationale

Low effort, high impact: turns recurring maintainer review comments into a failing CI signal. Prevents "skill merged but invisible in docs" drift as the registry grows (#148 wellness, future skills).

Complements #160 (bundle test presence) and #178 (example script correctness) without overlapping — this issue is index completeness, not tool-name dispatch or provider snippet content.

Implementation Idea

• New file: tests/test_registry_docs.py
• Use regex or simple markdown table parsing (no heavy deps).
• Parametrize failures with clear messages: "manifest wellness/mental_coach has no docs/skills/README.md row".
• Link from docs/TESTING.md § Framework tests.
• Optional follow-up: wire into [Docs]: Close out testing architecture epic #156 — status, checklist, and project board #179 testing epic checklist.
• Out of scope: validating Usage Example content ([Feat]: Automated check — skill catalog pages include all five provider Usage Examples #104), parsing examples/*.py for tool names ([Docs]: Audit runnable examples for manifest-derived tool name dispatch #178).

[Areen-09]

Hi! I'm a first-time contributor to this project and would love to take up this issue. Could you assign it to me?

[rosspeili]

Hey @Areen-09, welcome and you're assigned. Please make sure you read contributing guidelines as well as code of conduct and the AI native workflow documentation for your coding assistants. 🙏 I will review the PR once you open it. Ensure checklist and coherence before opening a PR. I would also suggest reading some of the merged PRs or feed them to your model for references. Good luck!