[Haoming9527]

What was wrong

Closes #380 (the requested documentation scope). Refs #346 and #367.

The documentation did not explain why an inherited-tool whitelist can leave the session's own ops_bash visible, or separate missing discovery configuration from a publication stall and a provider tool-change serialization rejection.

Change

Add paired English/Chinese troubleshooting guides and link them from usage. Document the owning shell row's publishBashTool: false switch, preserving other config; identify the host's inherited-versus-owned restriction boundary. Preset-only mounting is explicitly not presented as a verified isolation guarantee.

Explain that DSH/provider adapters own tool_addition message roles, with a bounded fresh-session/static-configuration diagnostic. Add effective-profile/discovery/presentation checks and crosslinks to #383, #390 and #395. The enabled-row stall described by #346 remains an evidence boundary; the guide does not claim that missing-row checks repair it or reproduce Desktop recovery.

Validation

• Both language guides' local links resolve.
• Both YAML examples parse through the repository's YAML parser and set the owning row's publishBashTool to false.
• git diff --check passed.
• The statements were checked against mountSessionBash, the pinned host's tool view/restriction implementation, actual host metadata/patch APIs, and the issue reports. No runtime implementation changed; no new tests or dependencies.
• Pull-request CI is green, including Linux, macOS, Windows, Node floor, source-size, tools trial, workflow lint and the required aggregate: https://github.com/T-Auto/dsh-ops/actions/runs/40112861899.

Environment

Documentation-only change, based on main 662a388. It does not alter an installed profile or send provider requests.

[Thatgfsj]

复审无误，批准合并