[ivankuznetsov]

Patrol Fix

This pull request repairs one finding tracked by Hive's Patrol Fix workflow.

Sources

• architecture_patrol: scheduled-68a9c8e760715d08694bc0d42d93ebc1e35e48d77b46f863e2c86c27c7ecbab9:centralize-brainstorm-document-location

Finding evidence

• {"file":"lib/hive/brainstorm_parser.rb","snippet":"Question = Struct.new(\n :round, :n, :text, :answer, :answer_encoding,\n keyword_init: true\n )","claim":"the canonical parser returns logical question data without source positions or block spans, so mutation callers cannot use its parse result to locate the corresponding document region"}
• {"file":"lib/hive/bot/brainstorm_answer_writer.rb","snippet":"def target_question_line_index(lines, parsed, question_n)\n matching = parsed.select { |q| q.n == question_n }\n target_position = matching.find_index { |q| q.answer.nil? }","claim":"the writer reconstructs physical identity from raw lines after parsing because the parser boundary does not expose that decision"}
• {"file":"lib/hive/bot/brainstorm_answer_writer.rb","snippet":"Previously this was a two-step scan: strict by-number first,\n # then by-position fallback. The strict scan ignored Q-context\n # and could cross round boundaries","claim":"the duplicated location rules have an observed behavioral consequence: an earlier implementation crossed round boundaries and routed a later-round answer into the wrong slot"}
• {"file":"lib/hive/brainstorm_parser.rb","snippet":"The writer's find_empty_answer_slot has a\n # matching lenient fallback so the answer lands in the\n # actual slot on disk.","claim":"the current parser explicitly depends on a matching writer-side structural rule, making future format changes a synchronized cross-boundary edit"}

Independent review

The patch correctly centralizes question and answer locations in Hive::BrainstormParser and makes both writer paths consume the parser's normalized line-index space. No patch-related defect was found. The controller's only red validation is an unrelated AgentCliRuntime timeout test; that exact test passed independently, while every focused and adjacent parser/writer check remained green.

• Reviewed the exact c2e6b77..07a5ee6 diff: only parser/writer logic, focused tests, and required wiki documentation changed; git diff --check was clean.
• Parser Question values now expose question_line_index, answer_line_index, and exclusive block_end_index from document_lines; append and ordinal writes use those fields for fill or insertion, eliminating writer-side rescans while preserving Q, Round, marker, EOF, CRLF, and lone-CR behavior.
• Independent focused verification passed: parser, writer, round-trip, and skill-contract tests reported 28 runs and 73 assertions; answer CLI and adjacent consumers reported 33 runs and 224 assertions; RuboCop inspected four changed Ruby and test files with no offenses.
• The controller's architecture test failed only in components/agent-cli-runtime/test/probe_test.rb, outside this six-file patch; rerunning that exact timeout test passed with 1 run and 5 assertions.

Validation

Verdict: passed

• Required CI: run 37165702708 passed on unchanged head 01aa23d389e0080e5b54365489e04dc67e9be9df.
• architecture:test: exit 1 (unrelated AgentCliRuntime timeout; exact retry passed)
• agent:parser-location-fields-and-answer-writer-round-trip: exit 0
• agent:answer-cli-and-adjacent-consumers: exit 0
• agent:style-gate-on-changed-files: exit 0

[ivankuznetsov]

Merged as part of #1535, which batched the remaining ready PRs (including this PR's head) into one CI run.