From omni
Aggregates omni-execution-log.json into omni-observability-result.json from feature directories and inserts into ssd_step_info database using Python scripts. Use when execution logs already exist.
npx claudepluginhub zte-aicloud/co-omnispec --plugin omniThis skill uses the workspace's default tool permissions.
- 某特性目录下**已经存在** `omni-execution-log.json`(根为步骤对象数组)。
Records execution metrics of skills (start/end time, duration, input/output, result) to JSON log at FEATURE_DIR/.runs/metrics/omni-metrics-log.json. Invoke after skill runs for tracking.
Loads structured data like logs, test results, and errors into SQLite for SQL queries, aggregations, and correlations. Replaces custom jq/Python parsing for >100-record datasets with multiple analyses.
Logs data analysis operations in JSONL journals with SHA256 hashes, metadata, and tool params for independent end-to-end reproducibility.
Share bugs, ideas, or general feedback.
omni-execution-log.json(根为步骤对象数组)。omni-observability-result.json(按 sdd_step / step_name 取最后一次执行),并写入数据库(或仅校验)。在 Cursor / Claude Code 中任选一种即可:
observability-publish),并说明特性目录或 omni-execution-log.json 的完整路径;若仅校验数据库,写明 --dry-run(仅对入库步骤生效)。/observability-publish,并按提示补充路径或 FEATURE_DIR。changes/002-xxx 执行 observability-publish:生成可观测性 JSON 并入库」。由 Agent 读取本文件后,在终端执行下方脚本,禁止只口述不执行(除非用户环境禁网禁库)。
.claude/skills/observability-publish/scripts/python/(相对仓库根目录)。{FEATURE_DIR}/omni-execution-log.json。{FEATURE_DIR}/omni-observability-result.json(由聚合脚本写在与输入同目录)。FEATURE_DIR 可为环境变量,或用户给出的 changes/<feature-id>-... 绝对/相对路径。
在仓库根目录执行(按实际路径替换):
python .claude/skills/observability-publish/scripts/python/omni_build_observability_result.py -i "<FEATURE_DIR>/omni-execution-log.json"
可选:
-o "<FEATURE_DIR>/omni-observability-result.json"(通常省略,默认即此名)。branch 时补足根级分支:--default-branch "<特性目录名或分支名>"。成功后应得到 omni-observability-result.json,且含 step_results、summary、invalid_records(若有无效行)。
未要求试跑时执行:
python .claude/skills/observability-publish/scripts/python/ssd_step_insert.py -i "<FEATURE_DIR>/omni-observability-result.json"
若步骤数据中可能缺 branch,且根对象也无 branch / feature_desc,需传入与上一步相同的
--default-branch。
用户要求不写库、只校验时,在本命令上加 --dry-run:
python .claude/skills/observability-publish/scripts/python/ssd_step_insert.py -i "<FEATURE_DIR>/omni-observability-result.json" --dry-run
注意:--dry-run 仅作用于 ssd_step_insert;聚合脚本仍会写出 omni-observability-result.json。若用户希望连 JSON 都不覆盖,应先备份或改用 -o 写到临时路径,再对该路径 dry-run 入库。
已写入: ... 与 插入成功,条数: N 或 [dry-run] 将插入 N 条。omni-observability-result.json 查看 summary 与 step_results 是否符合预期。pymysql 与 ssd_step_insert.py 内 DB_CONFIG;连接失败应向用户说明并保留已生成的 JSON。changes/<feature>/execution-observability-skill-plan.md(若存在)。