fix(security): preserve leading BOM in strip_dangerous

[dadavidtseng]

Summary

ContentScanner.strip_dangerous() was stripping all BOM characters (U+FEFF), including a leading BOM at position 0. This contradicts its documented contract:

Info-level characters (emoji selectors, non-breaking spaces, unusual whitespace) are preserved — they are legitimate and stripping them would break content.

scan_text() correctly classifies a leading BOM as "info" severity (standard practice for UTF-8 files), but strip_dangerous() unconditionally stripped it anyway. When apm audit --strip runs on a file with a legitimate leading BOM, the BOM would be incorrectly removed.

Fix

• Check whether the BOM is at position 0 before deciding to strip it
• Leading BOM (info-level) is now preserved; mid-file BOMs (warning-level) are still stripped
• Updated the corresponding test to assert the leading BOM is preserved

Files changed

• src/apm_cli/security/content_scanner.py — 3-line fix in strip_dangerous()
• tests/unit/test_content_scanner.py — updated test to match corrected behavior

Test plan

• All 78 content scanner tests pass
• All 41 audit command tests pass
• Leading BOM at position 0 is preserved (info-level)
• Mid-file BOM is still stripped (warning-level)

[dadavidtseng]

@microsoft-github-policy-service agree

[Copilot]

Pull request overview

This PR fixes ContentScanner.strip_dangerous() to preserve a legitimate leading BOM (U+FEFF) while still stripping mid-file BOMs, aligning behavior with the documented “info-level characters are preserved” contract.

Changes:

• Update strip_dangerous() to keep BOM at index 0 and remove BOMs elsewhere.
• Update the unit test to assert a leading BOM remains unchanged.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
src/apm_cli/security/content_scanner.py	Adjust BOM handling in strip_dangerous() to preserve the leading BOM only.
tests/unit/test_content_scanner.py	Update test to validate preservation of a leading BOM.

[danielmeppiel]

Clean, well-scoped fix. strip_dangerous() now correctly preserves leading BOM (info-level) while still stripping mid-file BOM (warning-level), aligning with the documented security model contract.

Please ensure CI tests pass before merging. Thanks for the contribution! 🎉

[dadavidtseng]

@danielmeppiel All CI checks have passed, but the Integration Tests (PR) required check is stuck at "Waiting for status to be reported" and was never triggered. Could you re-run it or advise on how to trigger it? Thanks!

[danielmeppiel]

It's a workflow gated on approval - it's running now, will merge. Thanks a lot for your contribution!