fast-filesystem-ops

Use fast_search_files to locate the function, then fast_read_multiple_files with a line range to load only that section, and finally edit_file with the target line numbers and replacement code. This avoids loading thousands of irrelevant lines.

Search all occurrences with fast_search_files filtered by language, read the relevant sections of each file in one batch with fast_read_multiple_files, then apply edit_file sequentially to each occurrence using search-and-replace mode.

For large JSON files, use json_query_jsonpath to extract the specific node you need instead of loading the entire file. Combine this with targeted edits only after confirming the exact location of the value to change.

Reading with a few lines of context prevents edit errors caused by misidentified boundaries. The skill recommends adding 2-3 context lines around target ranges so replacements land on the correct code without breaking syntax.

Avoid full file reads whenever a file exceeds roughly 1000 lines or when you only need a specific function or pattern. Always search first, then read only the matching sections to keep token usage under control.