feat: add standalone agent health check script

[huangrichao2020]

We received your feedback on the previous PR, and it was reasonable. I reworked the change accordingly and kept this follow-up much smaller and more focused.

This PR only includes a standalone audit helper, without modifying core runtime behavior.

Scope in this PR:

• add a standalone scripts/agent_health_check.py
• add tests for the standalone script
• add minimal README usage notes

Explicitly not included:

• no monkey-patching in ga.py
• no tool registration in assets/tools_schema.json
• no new or modified runtime memory files under memory/
• no unrelated macOS installer changes

Why a standalone third-party style matters

My intention here is to keep agchk usable as an external package or helper, rather than making it part of GenericAgent's core runtime.

That means users can adopt it in low-risk ways:

• install agchk separately and run it from the terminal
• call the standalone script manually when they want an audit
• ask the agent to run the script or shell command on demand

Instead of binding it into the agent loop, tools schema, or memory files, the idea is to keep it as an optional auditor.

How users would use it

Direct usage from the terminal:

python scripts/agent_health_check.py --target-dir /path/to/GenericAgent
python scripts/agent_health_check.py --mode tools --json

Typical day-to-day requests a user might give in chat:

• "Run a health check on the current GenericAgent checkout."
• "Audit only the tools layer and show me the highest-severity findings."
• "Check whether memory or prompt layers look bloated or duplicated."
• "Run the standalone audit script and summarize only the actionable issues."

So the intended pattern is:

• user asks for an audit in normal conversation
• the agent runs the standalone script or terminal command
• the audit stays external to core runtime behavior

Validation

Validation run:

• python3 -m py_compile scripts/agent_health_check.py
• uv run --with pytest python -m pytest tests/test_agent_health_check.py -q
• python3 scripts/agent_health_check.py --target-dir /Users/tingchim2pro/Desktop/GenericAgent --mode tools --json

Note: I also checked a broader local test run and hit pre-existing MiniMax test failures unrelated to this script, so I kept the verification focused on the new standalone helper.

[huangrichao2020]

We received your feedback on the previous PR and reworked this follow-up accordingly. The scope is now intentionally limited to a standalone audit helper plus tests/docs, without touching core runtime integration, runtime memory files, or unrelated installer changes.\n\nThe intended usage is also external-first: users can run the script manually, or in day-to-day conversation ask the agent to run the standalone audit command and summarize the result. In other words, the audit stays outside GenericAgent's core runtime path rather than becoming a built-in tool.