AI Guides › Skills & Agents
When To Build A Skill, And When You're Just Procrastinating On The Actual Task
By Nigel Guy · 3 min read
There's a particular kind of stalling that feels like progress: instead of doing the annoying task in front of you, you start building a skill to do it for you next time. An hour later you have a half-written procedure, a fresher sense of having "worked on it," and the original task is still sitting there, untouched, because building the tool was more interesting than using it.
The rule: build the skill after you've done the task by hand at least once this cycle, not instead of doing it — if the task is still undone, you don't have a skill problem yet, you have a doing-the-task problem.
The mechanism
- Do the task the plain way first, today. Not as a formality — because doing it once, without the skill, is how you learn what the actual steps are. A skill written before you've done the task is a guess at a procedure, not a procedure.
- Ask what's actually being avoided. Notice whether the urge to build tooling shows up specifically when the task is unpleasant, unclear, or has an awkward first step. That pattern is the tell. A task you enjoy rarely triggers the "let me automate this instead" impulse.
- Check the repeat count before you build anything. One completed instance of the task tells you what the steps are. It doesn't tell you the task recurs often enough to justify writing, testing, and maintaining a skill for it. Wait for the second or third time before you build.
- Time-box the build, separately from the task. If you do decide to build the skill, do it after the task is done and on its own slot of time — not as a replacement activity mid-task. That keeps the build from quietly consuming the time you meant to spend finishing the thing.
- Notice if the skill you're planning is more elaborate than the task warrants. A five-minute task doesn't need a skill with branching logic and edge-case handling. If the build is bigger than the problem, that's a second sign you're avoiding something, not solving it.
What to skip
Skip building a skill for a task you haven't actually completed yet by hand — you don't know the real steps, only the steps you imagine. And skip treating "I built a tool for it" as equivalent to "I did it." A skill sitting unused while the underlying task is still overdue hasn't saved you any time; it's cost you time you could see if you added it up honestly.
Guardrails
- This isn't an argument against building skills early — it's specifically about the case where skill-building is substituting for a task you're avoiding. If you've already done the task and are building ahead of the next repeat, that's normal and sensible planning.
- Procrastination dressed as productivity is hard to self-diagnose in the moment. If you notice you're building tooling and can't remember the last time you actually finished the underlying task without the tool, that's worth sitting with rather than pushing past.
- Nothing here tells you how often is "often enough" to justify a skill — that threshold is genuinely personal and task-dependent, and it's covered directly in "The Skill That Pays For Itself In One Use vs The One That Never Will."
All 751 AI guides · JulieMango plans from £17/mo