一键导入
agent-store-pipelines
Shell composition patterns for importing, exporting, and transforming agent-store records.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Shell composition patterns for importing, exporting, and transforming agent-store records.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
| name | agent-store-pipelines |
| description | Shell composition patterns for importing, exporting, and transforming agent-store records. |
agent-store is designed to compose with ordinary shell tools.
Batch create from lines:
while IFS= read -r line; do
agent-store create note text="$line"
done < notes.txt
Bulk import JSONL with create --stdin (one {"kind":...,"fields":{...}}
object per line, the find --json record shape; extra keys like id and
timestamps are ignored, so exports round-trip):
agent-store find kind=task --json | jq -c '.records[]' | agent-store create --stdin
Every line is validated before any record is created; an invalid line exits non-zero naming the line number with nothing imported.
Filter and format: --json list output wraps records in a
{"records":[...]} envelope, so iterate with .records[]:
agent-store find 'kind=task and status=pending' --json | jq -r '.records[].id'
JSON records include created_at and updated_at timestamps. Prefer the
built-in --sort, --desc, --limit, and --count flags over shell-side
sort, head, or wc -l:
agent-store find kind=log --sort updated_at --desc --limit 10
agent-store find kind=note --count
Capture command output:
agent-store create log command=test output="$(cargo test 2>&1)"