What this skill does, in one sentence
The “weekly-content-brief” Claude skill takes 5 inputs (focus topic, audience, primary keyword, word count, deadline) and outputs a complete Sunday content brief: 5 article titles, primary + secondary keywords, target word counts, internal link suggestions, and a quality checklist. I built it in 5 minutes on a Saturday afternoon. It saves me 3 hours every Sunday because I no longer write briefs from scratch. Total time invested: 5 minutes. Total time saved per year: 156 hours. ROI: 1,872x.
This is the smallest, dumbest, most leveraged Claude skill I’ve ever built. Here’s the exact code.
Why this works
Most “AI productivity” content tells you to write 400-line custom GPTs. That’s overkill for 80% of weekly tasks. The actual leverage is in tiny, repeatable scaffolds that Claude executes the same way every time.
A Claude skill is just a folder with two files: manifest.json and instructions.md. That’s it. The whole thing fits in a tweet.
The file structure
~/.claude/skills/weekly-content-brief/
├── manifest.json
└── instructions.mdTwo files. No build step. No dependencies.
manifest.json
{
"name": "weekly-content-brief",
"version": "1.0.0",
"description": "Generate a Sunday content brief for the upcoming week",
"tools": [],
"trigger_phrases": [
"weekly brief",
"content brief"
]
}That’s the entire manifest. The trigger_phrases array tells Claude when to load this skill automatically.
instructions.md (the actual logic)
# Weekly Content Brief Generator
When the user asks for a weekly content brief, follow these exact steps.
## Step 1: ask 5 questions in one message
Ask:
1. Focus topic for the week (e.g., "Claude skills", "newsletter monetization")
2. Primary audience (e.g., "solo founders", "marketing leaders")
3. Primary keyword to target
4. Total word count budget for the week (e.g., 8,000)
5. Publish deadline (e.g., "next Friday")
Wait for all 5 answers before proceeding.
## Step 2: produce the brief
Output exactly this format:
### Week of [date range]
**Focus:** [focus topic]
**Audience:** [audience]
**Primary keyword:** [keyword]
**Total budget:** [word count]
**Deadline:** [deadline]
### 5 articles to ship
For each article, output:
- **Title** (under 60 chars, includes a primary or secondary keyword)
- **Format** (tutorial / listicle / comparatif / case-study / tool-review)
- **Word count** (1,000-2,000 typical)
- **Primary keyword**
- **Secondary keywords** (3-5)
- **Internal link targets** (2-3 existing articles)
- **Hook angle** (1 sentence)
- **Why this article** (1 sentence justifying the topic)
### Pre-publish checklist for each
- [ ] Definition Lead in first 200 tokens
- [ ] At least 5 specific numbers (prices, percentages, durations)
- [ ] One comparison table OR numbered listicle
- [ ] FAQ with 5-7 questions
- [ ] 3 internal links
- [ ] Schema-valid frontmatter
## Step 3: stop
Do not generate the actual articles. Stop after the brief.That’s the full skill. 47 lines including blank lines.
How to install in 5 minutes
mkdir -p ~/.claude/skills/weekly-content-brief
cd ~/.claude/skills/weekly-content-briefCreate the two files above. Save them. Restart Claude Code. Done.
To use:
weekly brief pleaseClaude will ask the 5 questions, then output a complete brief. Total time per use: 4-6 minutes. Total time saved vs writing the brief from scratch: 35-50 minutes per use.
The math that justifies it
| Variable | Value |
|---|---|
| Time to build | 5 minutes |
| Time saved per use | 35-50 minutes |
| Frequency | Once per week |
| Time saved per year | 30-43 hours |
| Time saved if Maxime had a junior assistant doing this | $1,200-1,700/year at $40/hr |
| ROI multiplier | 600-1,000x |
This is the dumbest, smallest skill I’ve shipped. It’s also the most leveraged. The lesson generalizes.
Why most founders don’t build skills like this
“Most founders skip 5-minute skills because they look ‘too small to matter.’ They’re the ones that compound. The 4-hour custom skill pays for itself in 3 months. The 5-minute skill pays for itself the same week.”
Three reasons people skip these:
- Looks too simple to be valuable. It is simple. It still saves 3 hours/week.
- Feels like cheating. Building a “skill” sounds like you should write a 200-line YAML manifest. The trigger-phrases-and-prompt format is enough.
- No GitHub stars, no validation. These don’t go viral. They sit in your local config and quietly print value.
The pattern works for any repetitive task with a stable structure: weekly briefs, monthly P&L summaries, customer interview prep, sponsorship reply drafts.
4 more skills you can build the same way in 5 minutes each
| Skill | What it does | Time saved/week |
|---|---|---|
daily-standup | Asks you 4 questions, outputs a structured standup | 15-20 min |
customer-call-prep | Takes a customer name, outputs a 6-section pre-call brief | 30 min |
monthly-pl-summary | Takes 4 numbers, outputs a 1-page financial summary | 45 min |
sponsorship-reply | Takes inbound email + your rate, drafts a structured reply | 25 min |
Build all 4 in 25 minutes total. Save 2+ hours/week forever.
The trap to avoid
Don’t add tools (function calls, MCP, file I/O) to a skill that doesn’t need them. The point of these micro-skills is that they’re pure prompt scaffolding. The moment you add an external API call, your install time goes from 5 minutes to 90 minutes and the failure modes multiply.
If your skill needs a tool, it’s a different category. Build it. Just don’t pretend it’s still a 5-minute skill.
What I’d skip
I would not build these as skills:
- A skill that just rephrases what Claude already does well (e.g., “summarize this article” — Claude does this fine without a skill).
- A skill with more than 5 input questions (cognitive overhead defeats the time savings).
- A skill you’d use less than once a week (not worth the install).
FAQ
What’s the minimum viable Claude skill?
The minimum viable Claude skill is two files: a manifest.json with name, version, description, and trigger phrases, plus an instructions.md describing what Claude should do. Total minimum size: roughly 30 lines. No code, no API integrations needed.
Where do skills live on my computer?
Claude Code looks for skills in ~/.claude/skills/ by default. Each skill is its own subdirectory. You can have unlimited skills installed locally. They don’t slow down Claude — only the matched skill loads when triggered.
Can I share my skill with other founders?
Yes. Push the skill folder to a public GitHub repo, then anyone can clone it into their ~/.claude/skills/ folder. There’s no central registry yet, but the GitHub topic claude-skill lists 400+ community skills as of May 2026.
Do skills work in Claude.ai or only Claude Code?
This pattern (manifest.json + instructions.md) is specific to Claude Code. Claude.ai has Projects, which serve a similar purpose but use a different format (custom instructions + uploaded knowledge base files). The logic transfers, the file format doesn’t.
How do trigger phrases work?
Trigger phrases tell Claude Code when to auto-load a skill. If a user types a phrase containing one of the trigger words, Claude loads the skill’s instructions before generating its response. You can also load skills manually with /skill weekly-content-brief.
What’s a good first skill to build?
Build a skill for the task you do at the same time every week with the same structure. For most founders, that’s a weekly review, a content brief, or a customer prep. Pick the one you procrastinate on most. Building the skill is the procrastination cure.