I Replaced My Morning Tab Ritual With a Markdown File
I used to start every morning the same way. Open Jira, check the SRE support queue. Switch to GitHub, scroll through open PRs. Open Confluence, see if anyone updated the runbooks. Four browser tabs, ten minutes of clicking, and I still walked into standup feeling like I missed something.
Now I type /daily-routine in my terminal and get the full picture in about fifteen seconds.
Skills are just markdown files
Claude Code has this concept called skills. A skill is a markdown file you drop into .claude/skills/your-skill-name/SKILL.md. You give it a name, a description in the YAML frontmatter, and write instructions in the body. That’s the whole thing. No plugin SDK, no build step, no config files.
When you type the skill name as a slash command, Claude reads the file and does what it says.
I was skeptical at first. A markdown file felt too simple to be useful. But simplicity turned out to be the point.
What my morning routine actually does
The skill pulls from three sources:
Jira, through the Atlassian MCP server. I have two JQL queries — one for the SRE support request queue, one for Platform Engineering tickets updated recently. The MCP server handles authentication, so the skill file just contains the query and a reference to our Atlassian instance.
Confluence, also through MCP. It searches for recently edited pages in the team’s spaces. Useful for catching runbook updates or postmortem drafts I might have missed.
GitHub, through the gh CLI. Lists open PRs on our main infrastructure repo with review status.
There’s one small trick that makes Monday mornings less painful: the time window adjusts automatically. On Monday it looks back three days to cover the weekend. Other days, just yesterday.
The output lands as a structured summary. My open items first, then team activity grouped by status, then PRs and docs. Enough context to have a real conversation in standup instead of stalling while I frantically search Jira.
MCP makes this actually work
The part that surprised me was how well MCP tools fit into this pattern. The Atlassian MCP server gives Claude direct access to Jira and Confluence APIs. In the skill file, I just describe what I want queried. Claude calls the tool, gets structured data back, and formats it.
No curl commands. No token management. No OAuth refresh logic. The MCP server is a process running locally that handles all of that.
If you haven’t tried MCP servers with Claude Code yet, this kind of workflow is where they really pay off. Writing raw API calls in a skill would work but it’d be brittle. MCP abstracts the annoying parts.
What I’d change
The skill currently checks one GitHub repo. We have work spread across three or four repos, so I’ll extend it. That’s literally adding a few more lines to the markdown.
I also thought about wrapping the whole thing in a shell script, but I couldn’t find a reason to. The MCP tools handle Atlassian, gh handles GitHub, and Claude handles the formatting and summarization. Adding a script would just be one more thing to maintain for no benefit.
The actual productivity win isn’t code generation
Most conversations about AI developer tools focus on writing code faster. Autocomplete, function generation, test scaffolding. Fine, that stuff helps. But it’s not where I lose the most time.
The real drain is context gathering. Figuring out what happened since yesterday. What’s blocked. What needs attention before the day gets away from me. That work is invisible and repetitive and nobody ever optimizes for it.
A markdown file that assembles that context should save me ten to fifteen minutes every morning. And unlike generated code, there’s nothing to review for correctness. Either the Jira tickets are there or they’re not.
If you use Claude Code and you have some repetitive workflow you run through every day or every week, try building a skill for it. Start with something small. The barrier is writing a markdown file, and the payoff is immediate.