[Docs]: Audit runnable examples for manifest-derived tool name dispatch

[rosspeili]

What needs to be fixed?

#173 / PR #175 fix gemini_wallet_check.py and claude_wallet_check.py for finance/wallet_screening. Several other runnable examples still hardcode tool names that may not match bundle["manifest"]["name"] or sanitized adapter names — the same silent dispatch failure mode.

Audit and update examples to prefer TOOL_NAME = bundle["manifest"]["name"] (Gemini/Claude) or openai_tool["function"]["name"] (OpenAI/DeepSeek), matching gemini_evm_tx_handler.py / claude_evm_tx_handler.py.

Known hardcoded comparisons on main:

• examples/claude_pdf_form_filler.py → pdf_form_filler
• examples/gemini_pdf_form_filler.py → pdf_form_filler
• examples/gemini_wallet_check.py / claude_wallet_check.py → fixed in docs(wallet_screening): align examples and docs with finance/wallet_s… #175 (verify after merge)

Out of scope: loader/adapter behavior; manifest renames (separate RFC).

Acceptance:

• All examples/* agent-loop scripts dispatch via manifest-derived or adapter-derived names
• examples/README.md unchanged unless scripts renamed
• CHANGELOG [Unreleased] entry

Affected Page

examples/claude_pdf_form_filler.py, examples/gemini_pdf_form_filler.py, examples/gemini_wallet_check.py, examples/claude_wallet_check.py, docs/usage/agent_loops.md (if cross-links needed)

[rosspeili]

Please also cover docs/usage/agent_loops.md in the same PR (or a tight follow-up):

• Add missing row: defi/evm_tx_handler → gemini_evm_tx_handler.py, claude_evm_tx_handler.py
• After Add wellness/mental_coach skill resolving #148 #174 merges: wellness/mental_coach → mental_coach_demo.py (local execute)

Rule of thumb for new skills: if you add or change a script under examples/, update both examples/README.md and the agent_loops.md table.