Agent Skills
› Lifecycle-Innovations-Limited/claude-ops
› ops-research-audit
ops-research-audit
GitHub用于审计搜索和研究工具的可用性、配置及最佳实践。通过只读调用验证工具状态，对比官方文档生成包含现状、依据、影响和改进建议的审计报告，严禁修改配置或发送消息。
Trigger Scenarios
审计我的搜索工具
检查我的MCP工具
Install
npx skills add Lifecycle-Innovations-Limited/claude-ops --skill ops-research-audit -g -y
SKILL.md
Frontmatter
{
"name": "ops-research-audit",
"effort": "medium",
"maxTurns": 30,
"description": "OPS on-demand: This skill should be used when the user asks to \"audit my search tools\", \"check my MCP…",
"allowed-tools": [
"Read",
"Grep",
"Glob",
"WebSearch",
"WebFetch",
"Skill"
],
"argument-hint": "[tool-or-server]"
}
OPS ► RESEARCH AUDIT
Load ops-rules before acting. Public repo (no personal data). Outbound: one draft → one approval → one send. If AskUserQuestion / Workflow are missing, follow Rule 10 in ops-rules (Hermes: numbered options / two-turn Telegram card; delegate_task).
Read the research contract in full first. Its audit section (§5) defines this command; sections 1–4 govern every source you cite.
Steps
- Discovery/configuration: list the research and search tools this host offers in this session. Do not read shell configuration, dotfiles or secret stores to find more.
- Actual invocation: make one harmless read-only call per relevant tool. Record the real result or error. A listed tool or a handshake is not proof it works; a failure is a coverage gap.
- Best-practice evidence: read the official source for each tool you report on (contract §1, two-route cap §2).
- Report the three parts separately. Per finding: current state, official source (URL + passage), effect, smallest proposal, acceptance test.
Never install, change configuration, send outbound messages, rotate secrets or restart services as part of the audit. Proposals go to the operator.
Version History
- 6ac2237 Current 2026-10-18 17:54
