Decision making

Skill sandev/engineering-os/skills/strategy-and-communication/decision-making

An operating system for engineering: durable, opinionated playbooks (Agent Skills) for software architecture, AI/agentic systems, product engineering, and technical leadership. Read by humans, loaded by any LLM/agent.

Install

npx -y skills add sandev/engineering-os --skill decision-making

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Make and drive good decisions under uncertainty, clarifying the decision, who owns it, how reversible it is, and how to move forward without stalling. Use when a team is stuck, a choice has unclear ownership, a group can't reach consensus, or you need to decide with incomplete information.

SKILL.md

3.4 KB, as published. Nobody here has run it

Decision-Making

Most organizational drag is not bad decisions, it's undecided decisions: choices that stall because no one owns them, everyone waits for certainty that never comes, or the group confuses consensus with progress. The skill is deciding well and at the right speed, then committing.

Match speed to reversibility

The single most useful lens is one-way vs two-way doors:

Two-way door (reversible): decide fast, with whoever's closest to the problem. Bias to action; you can undo it. Waiting costs more than a wrong-and-corrected choice.
One-way door (irreversible or very costly to reverse): slow down, gather input, consider alternatives, and record the reasoning as an ADR (see the architecture-decision-records skill).

Most decisions are two-way doors treated, wrongly, as one-way. Misclassifying them is what makes orgs slow.

Clarify the decision before debating it

Stalled decisions are usually unclear decisions. Pin down:

What exactly are we deciding? State it as one question with a date.
Who owns it? One accountable decision-maker (a DACI/RACI driver), not "the group." Consensus is a nice-to-have, not a requirement.
What would change our mind? The information or threshold that resolves it, so you know when you have enough.
What's the default if we don't decide? (Often worse than either option.)

Decide with incomplete information

You will rarely have certainty; waiting for it is itself a (bad) decision.

Decide when you have ~70% of the information you wish you had; more than that and you're likely too slow (the Bezos heuristic).
Make the reversible call now and learn; reserve deliberation for the irreversible.
Separate facts, assumptions, and opinions explicitly, and name the assumptions you're betting on.
Beware bias: sunk cost (past spend isn't a reason to continue), confirmation (seek the disconfirming view), and HiPPO (loudest/highest-paid opinion isn't evidence).

Disagree and commit

Consensus is not required; alignment after the decision is. Solicit dissent before deciding (actively invite the strongest counter-argument) then, once the owner decides, everyone commits and executes as if it were their own, even those who disagreed. Re-litigating a made decision in the hallways is corrosive.

Write it down

Record what was decided, by whom, why, the alternatives, and what would trigger a revisit. This kills re-litigation, onboards newcomers, and lets you learn from outcomes. For strategy-level choices, see the technical-strategy skill; to communicate the decision upward, see the writing-for-executives skill.

Anti-patterns

Treating reversible decisions as if they were irreversible (analysis paralysis).
No clear owner, the decision belongs to "the group" and so to no one.
Waiting for certainty that will never arrive.
Requiring consensus, then getting the lowest-common-denominator choice.
Deciding by loudest voice or seniority rather than reasoning and evidence.
Reopening a settled decision every time someone's unhappy.
Never recording decisions, so the org re-debates them forever.