AgentSkillsCN

Research

针对内部资源（Confluence、Jira、ADO、PagerDuty、Azure、Outlook），开展文献资料、归属权、状态、基础设施及事件等方面的调研。

SKILL.md
--- frontmatter
description: "Research internal sources (Confluence, Jira, ADO, PagerDuty, Azure, Outlook) for documentation, ownership, status, infrastructure, and incidents"
allowed-tools: ["Task", "Skill", "Read", "Glob", "Grep"]
model: sonnet
context: fork

Internal Research

Research internal sources using parallel Task execution. ALWAYS wait for Task completion before proceeding.

Step 1: Dispatch Search Tasks (PARALLEL)

Launch relevant advisors in a single message - they run concurrently:

code
Task(subagent_type: "dataops-assistant:confluence-advisor", prompt: "{query}")  // docs, architecture
Task(subagent_type: "dataops-assistant:jira-advisor", prompt: "{query}")        // tickets, status
Task(subagent_type: "dataops-assistant:pagerduty-advisor", prompt: "{query}")   // incidents, on-call

For ADO (code, PRs, pipelines): Use bash scripts directly:

bash
~/.dataops-assistant/bin/ado-repos.sh <project> [filter]
~/.dataops-assistant/bin/ado-repo-activity.sh <project> <repo> [--since YYYY-MM-DD] [--author email]
~/.dataops-assistant/bin/ado-pipelines.sh <project> [filter]
~/.dataops-assistant/bin/ado-pipeline-runs.sh <project> <pipeline-id> [--top N]
~/.dataops-assistant/bin/ado-assigned-prs-fast.sh --project <project> [user] [--status active|completed|all] [--since YYYY-MM-DD]
~/.dataops-assistant/bin/ado-reviewed-prs-fast.sh --project <project> [user] [--status active|completed|all] [--since YYYY-MM-DD]

WAIT for all Tasks to complete before Step 2.

Step 2: Read Items (PARALLEL via Task)

From the reading plans, launch reader Tasks in a single message:

code
Task(subagent_type: "general-purpose", prompt: "Use /dataops-assistant:confluence-page-reader to read page {id}. Context: {query}")
Task(subagent_type: "general-purpose", prompt: "Use /dataops-assistant:jira-issue-reader to read {KEY}. Context: {query}")
Task(subagent_type: "dataops-assistant:ado-pr-reader", prompt: "{project}:{repo}:{pr} | {query}")

Available readers:

  • •confluence-page-reader skill - Confluence pages
  • •jira-issue-reader skill - Jira issues
  • •ado-build-reader skill - ADO builds
  • •ado-pr-reader agent - ADO pull requests (use agent directly, not via skill)
  • •read-macos-outlook skill - Outlook emails
  • •read-macos-outlook-calendar skill - Outlook calendar

WAIT for all Tasks to complete before Step 3.

Step 3: External Research (if needed, PARALLEL)

For external concepts, launch research Tasks:

code
Task(subagent_type: "general-purpose", prompt: "Use /dataops-assistant:openai-research to research: {topic}")

WAIT for completion.

Step 4: Synthesize

Only after ALL Tasks complete, output:

markdown
## Research: {Question}

### Answer
[Direct answer]

### Key Findings
1. [Finding + source]

### Sources
- [Links]

### Confidence
**{High/Medium/Low}** - [Why]

### Gaps
- [What's missing]