Result Aggregation

Result Aggregation

Goal

Aggregate priority execution records and defect records into root summary files.

Inputs

Read from:

执行结果/P0/
执行结果/P1/
执行结果/P2/
执行结果/P3/

Outputs

Generate or update at the module artifact root:

• 05-执行记录.md
• 06-缺陷记录.md

Regression Inputs

When present, also read priority-owned regression artifacts:

• 执行结果/Px/07-回归记录.md
• 执行结果/Px/08-回归缺陷状态.md
• 执行结果/Px/regression-screenshots/
• 执行结果/Px/regression-videos/
• 执行结果/Px/regression-traces/
• 执行结果/Px/regression-scripts/

Root summaries must keep first-run results and regression results distinguishable. Do not overwrite original first-run defects; record regression status as a status update.

Rules

• Root summaries are owned by the main agent.
• Read priority-level 05-执行记录.md and 06-缺陷记录.md.
• Treat 通过, 失败, 阻塞, 未测, and 待确认 as the only valid first-run case statuses. Do not silently map unknown or blank values; flag them for correction in the priority record.
• Summarize total, passed, failed, blocked, untested, and pending-confirmation cases from those five statuses. The total must equal their sum.
• Preserve the priority directory boundary and relative evidence paths. Priority agents own 执行结果/Px/; only the main agent writes root 05-执行记录.md and 06-缺陷记录.md.
• Carry application type, service status and ownership, page-readiness basis, locator type, Console/Network observation, and session/temporary-service cleanup outcomes into the root summary or its priority index. Do not erase a priority-level exception by replacing it with a batch-wide default.
• Keep an actual-result summary separate from diagnostic evidence. Passing cases need a concrete result summary but do not require screenshots or diagnostic artifacts.
• Include Console, Network, video, Trace, and HAR entries only when the execution tool actually produced them. Preserve 未产生（原因） when a priority record reports that material was not generated; never infer evidence from a backend's capabilities.
• For failed cases, preserve available failure evidence and any stated reason that expected evidence was not produced. Do not turn a missing optional diagnostic artifact into a successful observation.
• Mark child-agent write boundary violations.
• Do not convert blockers into defects without user confirmation.