What problem does it solve? Diagnosing file descriptor (FD) and handle leaks on HarmonyOS devices requires manually reading large diagnostic logs, correlating handle type distributions, directory clusters, per-resource details, and FdTrack allocation stacks. This Skill automates that analysis and produces a structured root-cause report with an auditable evidence chain. ## Core Features & Use Cases - Automated Log Parsing: A Python script extracts the fault snapshot (leaked fd nums), Leaked fd Top 10, Dir Type Top 10, five special-resource detail sections (ashmem, socket, pipe, sync_file, dmabuf), and FdTrack allocation stack hotspots from [pid]_fd_leak.txt or RESOURCE_OVERLIMIT_[TIMESTAMP]_[PID].log files, outputting Markdown or JSON. - Fault Mode Library Matching: Matches evidence against a built-in fault mode library covering two second-level root causes (in-process handle leak over 20000 handles, RS-side handles over 25000 with RENDER_MEMORY_OVER_WARNING) and their third-level causes such as file, ashmem, socket, dmabuf, pipe, eventpoll, and eventfd leaks. - Confidence-Graded Conclusions: Outputs a three-level root cause table, evidence chain, fix suggestions, and a HIGH/MEDIUM/LOW confidence rating while strictly distinguishing the leak snapshot from 10-minute FdTrack allocation statistics. - Use Case: A user provides a RESOURCE_OVERLIMIT_20240627115528_1380.log from a background app killed for resource overlimit; the Skill parses it, finds 1337 ashmem handles dominating the snapshot, cross-checks ashmem detail records and FdTrack stacks, and reports an ashmem handle leak with MEDIUM confidence plus symbolization advice. ## Quick Start Analyze the attached FD leak log file and produce a root-cause report with evidence chain and fix suggestions.