Context
Methodology analysis from an RLCR session (8 rounds, all ACs met). The session completed successfully but identified patterns that could reduce round count from 8 to 5-6 for similar complexity.
Improvement Suggestions
S1: Front-load negative-path enumeration in reviews (High Impact)
Pattern: Reviewer discovered negative-path gaps one layer at a time across multiple rounds, each time finding a deeper requirement within the same acceptance criterion.
Improvement: When a reviewer identifies a gap in an acceptance criterion, they should exhaustively enumerate all negative paths for that criterion in the same review, rather than reporting only the most obvious one.
Expected impact: 2-3 fewer rounds in sessions with complex validation requirements.
S2: Require regression tests for control-flow changes
Pattern: Two fixes introduced new regressions (validation bypass in fallback path, initial-prompt rejection in empty-base path). Both were control-flow changes where a new early-return or guard was added without considering all callers.
Improvement: Any fix modifying control flow (adding early returns, guards, or fallback paths) must include a regression test for the inverse case. Round contracts should explicitly list "regression risk areas" when the fix involves branching logic.
Expected impact: Fewer fix-introduces-regression cycles, saving 1-2 rounds per session.
S3: Batch code-review findings into a single round
Pattern: Post-AC code review rounds each fixed 1-2 findings. The reviewer issued findings one batch at a time, requiring a new round for each batch.
Improvement: After the AC-driven phase completes, the code reviewer should perform a single comprehensive review and report all findings at once. The implementer then addresses them in one round.
Expected impact: 1-2 fewer rounds in the post-AC phase.
S4: Plans should specify negative test cases explicitly (High Impact)
Pattern: Plans specified acceptance criteria at a behavioral level but did not list specific negative test cases. The implementer interpreted these narrowly, and the reviewer had to expand the interpretation across multiple rounds.
Improvement: Plans should include a "negative test matrix" for each acceptance criterion, listing specific invalid inputs and expected error types.
Expected impact: 1-2 fewer rounds for requirements with complex validation logic.
S5: Add a "regression risk" field to round contracts
Pattern: Round contracts listed objectives and success criteria but did not flag areas where the fix might break existing behavior.
Improvement: Each round contract should include a "regression risk" section identifying which existing behaviors could be affected by the planned changes.
Expected impact: Earlier detection of regressions, reducing fix-on-fix cycles.
S6: Compact the goal tracker for completed items
Pattern: The goal tracker accumulated 27 task entries across 10 plan versions, making it increasingly difficult to scan for active work.
Improvement: Once a task is verified, collapse it into a summary line. Keep detailed entries only for active and deferred tasks.
Expected impact: Better readability for long sessions, reducing cognitive overhead.
Priority Assessment
Highest impact: S1 (front-load negative paths) + S4 (negative test matrices in plans). These address the root cause of the "peeling the onion" pattern that drove most of the extra rounds.
Context
Methodology analysis from an RLCR session (8 rounds, all ACs met). The session completed successfully but identified patterns that could reduce round count from 8 to 5-6 for similar complexity.
Improvement Suggestions
S1: Front-load negative-path enumeration in reviews (High Impact)
Pattern: Reviewer discovered negative-path gaps one layer at a time across multiple rounds, each time finding a deeper requirement within the same acceptance criterion.
Improvement: When a reviewer identifies a gap in an acceptance criterion, they should exhaustively enumerate all negative paths for that criterion in the same review, rather than reporting only the most obvious one.
Expected impact: 2-3 fewer rounds in sessions with complex validation requirements.
S2: Require regression tests for control-flow changes
Pattern: Two fixes introduced new regressions (validation bypass in fallback path, initial-prompt rejection in empty-base path). Both were control-flow changes where a new early-return or guard was added without considering all callers.
Improvement: Any fix modifying control flow (adding early returns, guards, or fallback paths) must include a regression test for the inverse case. Round contracts should explicitly list "regression risk areas" when the fix involves branching logic.
Expected impact: Fewer fix-introduces-regression cycles, saving 1-2 rounds per session.
S3: Batch code-review findings into a single round
Pattern: Post-AC code review rounds each fixed 1-2 findings. The reviewer issued findings one batch at a time, requiring a new round for each batch.
Improvement: After the AC-driven phase completes, the code reviewer should perform a single comprehensive review and report all findings at once. The implementer then addresses them in one round.
Expected impact: 1-2 fewer rounds in the post-AC phase.
S4: Plans should specify negative test cases explicitly (High Impact)
Pattern: Plans specified acceptance criteria at a behavioral level but did not list specific negative test cases. The implementer interpreted these narrowly, and the reviewer had to expand the interpretation across multiple rounds.
Improvement: Plans should include a "negative test matrix" for each acceptance criterion, listing specific invalid inputs and expected error types.
Expected impact: 1-2 fewer rounds for requirements with complex validation logic.
S5: Add a "regression risk" field to round contracts
Pattern: Round contracts listed objectives and success criteria but did not flag areas where the fix might break existing behavior.
Improvement: Each round contract should include a "regression risk" section identifying which existing behaviors could be affected by the planned changes.
Expected impact: Earlier detection of regressions, reducing fix-on-fix cycles.
S6: Compact the goal tracker for completed items
Pattern: The goal tracker accumulated 27 task entries across 10 plan versions, making it increasingly difficult to scan for active work.
Improvement: Once a task is verified, collapse it into a summary line. Keep detailed entries only for active and deferred tasks.
Expected impact: Better readability for long sessions, reducing cognitive overhead.
Priority Assessment
Highest impact: S1 (front-load negative paths) + S4 (negative test matrices in plans). These address the root cause of the "peeling the onion" pattern that drove most of the extra rounds.