Skip to content

Firzus/agent-skills / skills/stack/interview

interview

Verified

Use when the user asks to be interviewed or to have a plan challenged, or when another skill hands over decisions the user owns. Not for choices the agent can take itself or that research or an experiment can settle.

Settle decisions by interview

Reach a shared understanding of what will be built, with every decision the user owns taken by the user and every other choice taken by you.

What to ask

A question is for the user only when its answer changes what gets built and the user owns it: behavior, scope, cost, what happens when things go wrong, or a step that is hard to undo. A choice that an experiment or research could settle is a need for the caller, not a question.

Before asking, walk through the normal case and the cases that matter for this work: empty or invalid input, failure halfway, repetition, two people acting at once, existing data, and who is allowed to do what. Ask only where the answer is not already evident.

Rounds

  1. Ask every independent question in one round, most consequential first. Keep a question that depends on an unanswered one for a later round.
  2. "You choose" makes the decision yours, recorded with its reason. Ask the follow-up questions each answer opens.
  3. When nothing consequential remains open, summarize the decisions, including yours, and what stays out of scope, and ask the user to confirm it. A caller may merge this summary into its own agreement message.

What each question contains

Write each question to the decision format of the operating policy, with each option on its own line and lettered so the user can answer briefly (such as 1b). Ask in a plain message, never through a question tool the harness provides: its short labels and limited options cannot carry each option's gains, costs, and reversibility, nor the whole round in one place.

Return

Return the accepted decisions with their consequences and who took each, the exclusions, open questions, needs for research or an experiment, and out-of-scope findings kept apart.

Done: the user confirmed the summary. Waiting: a question or the confirmation is unanswered. Stopped: the user ended it; return the partial result.

reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall