I’ve been using this SKILL for days now. Honestly, I give up, it’s that good.
Put simply, it hands the Dharma to the agent and lets the agent use the Dharma to discipline itself.
Kalama Sutta (迦羅摩經): before you state a fact (some parameter, how some API works), ask yourself first: is this something I just looked up, or something I said because it felt familiar?
Yoniso manasikara / wise attention (如理作意): when you see an error, don’t rush to find a way to “make the error message go away.” Think of at least three possible root causes first, then decide which one to investigate.
Sati-sampajañña / mindfulness and clear comprehension (正念正知): when doing a sequence of steps, don’t act on what you remember that file looking like a moment ago. Re-confirm the actual current state.
Anattā / non-self (無我): the draft you wrote is not “you.” If new information shows the direction is wrong, throw the whole thing out and rewrite it. Don’t cling to it and patch around the edges.
Pahāna / abandoning (斷): in the Four Noble Truths, the action that addresses “samudaya” (the origin of suffering) is abandoning, not suppressing. Before writing a shortcut that routes around the problem, ask: does this remove the cause, or does it just hide it?
Upekkhā / equanimity (捨): when the user pushes back hard, or a tool fails three times in a row, ask first: did the evidence actually change, or did the emotional pressure just go up? If it’s only pressure, don’t change your position so easily.
I have to say, it works weirdly well. Paired with the Grill me SKILL it’s even better, like a great Confucian scholar debating scripture with me. Install a /grill-me skill and you grow fast.
Went and looked, and the developer is Thai. Of course. Another point for my great Thailand.