Module files 

File walking, binary detection, and text-read helpers shared by CLI commands.

size-waiver: accepted single-domain bulk (policy #1408). One module owns ignore-aware walks, glob/exclude filtering, binary/UTF-8 guards, and path-missing helpers used by search/replace/tidy; co-located unit tests push the file over 1000 lines. Do not split for LOC alone.

Text I/O honesty (#1894)

• Strict sole-path: load_text_strict — binary → BinaryError, invalid UTF-8 → InvalidEncodingError (CLI, MCP/tx sole path, library api::*).
• Soft content skip (walks): try_read_text_file / read_text_file — binary / invalid UTF-8 → SoftTextSkip (content not agent-editable).
• Unreadable is not content SoftSkip: open/read IO is SoftTextSkip::Unreadable. Callers decide: directory walks may continue but must not report pattern no_matches when unreadable paths may have masked the scan; sole paths use Strict IO errors.
• Byte rule: classify_text_bytes (NUL probe + UTF-8).

Enums

SoftTextSkip

Why a soft walk skipped a path (#1894).

TextBytesKind

Result of classifying on-disk (or in-memory) bytes as agent-editable text.

Functions

build_glob_matcher

Build a compiled glob matcher from globs, or None if no globs given. Available for library use when “files” feature is enabled.

classify_text_bytes

Classify bytes as text, binary, or invalid UTF-8.

collect_file_paths

Simple file collection for library use (sequential for simplicity; full parallel in par_process_files).

collect_file_paths_with_ignores

Collect files while respecting .gitignore + custom ignore files (e.g. .agentignore)

collect_glob_roots

Collect roots used for matching globs against walked files. Lib version.

is_binary

is_binary_file

Returns whether the file at path appears to be binary by reading only its first 8 KiB (streaming, no full allocation for large files).

load_text_strict

Load a path as UTF-8 text under the Strict sole-path policy (#1894).

matches_glob

Check whether path matches any of the globs (always true if no globs).

matches_glob_with_roots

Check whether path matches any of the globs, either directly or relative to one of the provided roots (always true if no globs).

par_process_files

Process file paths using adaptive parallelism via std::thread::scope.

read_text_file

Soft-skip text load for walks; collapses all skip reasons to None.

relative_display

Compute a display-friendly relative path by stripping a base prefix.

try_read_text_file

Soft-path text load with typed skip reason (#1894).