| name | offeru |
| description | Use when operating OfferU through Claude Code or another external agent. Provides CLI-first discovery, atomic operations, safe workflow planning, and human-confirmed side-effect execution. |
OfferU Agent-Native Operating Skill
Work from backend/ unless the user asks for frontend or extension work.
First commands
Run these before controlling the product:
python -m app.cli doctor --pretty
python -m app.cli manifest --pretty
python -m app.cli run agent_playbook --arg detail=full --pretty
Use the returned JSON as the live source of truth. Do not guess API paths or operation parameters.
Control rules
- Prefer
python -m app.cli run <operation> over raw HTTP.
- One CLI call performs one atomic operation.
- Discover operations with
python -m app.cli ops --pretty.
- Inspect parameters with
python -m app.cli schema <operation> --pretty.
- Use
python -m app.cli run workflow_catalog --pretty to see built-in workflows.
- Use
python -m app.cli run workflow_plan --arg goal="<goal>" --pretty to generate a command sequence.
- Read operations execute directly.
- Operations with
write, llm, or external side effects must be dry-run first.
- Never auto-submit applications, send emails, or message external parties.
Core workflows
- Daily review:
workflow_plan --arg goal="今日岗位概览"
- Batch triage:
workflow_plan --arg goal="批量筛选岗位"
- Tailored resume:
workflow_plan --arg goal="定制简历"
- Application pipeline:
workflow_plan --arg goal="创建投递待办"
- Workspace handoff:
workflow_plan --arg goal="当前页面上下文接管"
Validation
After changing backend agent/CLI behavior, run:
python -m compileall app tests
python -m unittest tests.test_cli_ops -v