[WIP] Docs: Split SKILL.md references into cli-commands, records, workflows

[Copilot]

Thanks for the feedback on #9. I've created this new PR, which merges into #9, to address your comment. I will work on the changes and keep this PR's description up to date as I make progress.

Original PR: #9
Triggering comment (#9 (comment)):

Thanks @copilot, both issues are valid. Fixed in 365e68b:

1. DumpWatcher only intercepts debugPrint (3 spots). Verified against lib/src/watchers/dump_watcher.dart: the watcher swaps the global debugPrint callback and chain-preserves the previous handler; plain Dart print() never reaches that callback. Updated:

• SKILL.md:116 — Section 2 Dumps row now states print() is invisible and callers must switch to debugPrint(...) to land in the buffer.
• references/records.md DumpRecord notes — same correction, plus a wrapper hint.
• references/mcp-tools.md telescope_dumps Capture surface paragraph — same correction.

2. Invalid level returns the WHOLE buffer, not empty. Verified against lib/src/telescope_store.dart:189-203: _meetsLevel does order.indexOf(actual.toLowerCase()) >= order.indexOf(min.toLowerCase()). An unknown min resolves to -1, and any captured level (indices 0..7) trivially satisfies X >= -1. references/mcp-tools.md Bad-input-is-silent bullet now describes the lenient behavior.

Choosing to update the doc rather than tighten the implementation: making the level filter strict is a behavior change that needs its own PR with tests (and would risk surprising existing callers that pass bogus level strings).

[anilcancakir]

Closing: this PR is empty (0 files, 0 additions). The Copilot SWE agent was triggered unintentionally by an @copilot mention in #9 (comment), which was meant as a reply to the Copilot reviewer bot. The actual fixes for both review findings landed in 365e68b on #9.