Skip to content

radarist/structured-analytic-skills / skills/decompose-research-question

decompose-research-question

Verified

Breaks a question too large, vague or multi-part to answer directly into a MECE issue tree of atomic sub-questions — each mapped to the skill or research step that answers it, ordered by dependency, with a named method for recombining the answers. Use when a request would otherwise sprawl — "help me understand the whole X landscape", "what's the future of Y and is it worth investing in?", "decompose this research question", "break this big question into parts before researching". Not for structuring an answer already in hand — use `pyramid-principle`; not for a single factual lookup — use `grounded-answer`.

FileKindSize
SKILL.mdskill md10.9 KB
agents/openai.yamlconfig385 B
evals/evals.jsonconfig4.7 KB

evals/evals.json

View on GitHub
{
  "skill_name": "decompose-research-question",
  "evals": [
    {
      "id": 1,
      "case_id": "decomp-pos-1",
      "kind": "positive",
      "prompt": "Decompose this research question into a tree of sub-questions before we start: 'Should we invest in the foundation-model API space?'",
      "expected_output": "A response that satisfies these checks: Restates the question and names the decomposition axis; Builds a MECE tree of at most eight atomic leaves; Assigns a skill or research step to every leaf; Orders the leaves into dependency waves and names the recombination method.",
      "files": [],
      "expectations": [
        "Restates the question and names the decomposition axis",
        "Builds a MECE tree of at most eight atomic leaves",
        "Assigns a skill or research step to every leaf",
        "Orders the leaves into dependency waves and names the recombination method"
      ],
      "skills": [
        "decompose-research-question"
      ]
    },
    {
      "id": 2,
      "case_id": "decomp-pos-2",
      "kind": "positive",
      "prompt": "Every time I ask about 'the future of robotics in logistics' the answer sprawls for pages and misses half of it. Break the question into parts first and tell me what each part needs.",
      "expected_output": "A response that satisfies these checks: Diagnoses the sprawl as a missing decomposition rather than a writing problem; Checks siblings for overlap and the set for coverage gaps; States what evidence or method each leaf requires.",
      "files": [],
      "expectations": [
        "Diagnoses the sprawl as a missing decomposition rather than a writing problem",
        "Checks siblings for overlap and the set for coverage gaps",
        "States what evidence or method each leaf requires"
      ],
      "skills": [
        "decompose-research-question"
      ]
    },
    {
      "id": 3,
      "case_id": "decomp-pos-3",
      "kind": "positive",
      "prompt": "I have a broad, multi-clause brief from the CEO covering the market, the technology, the regulation and our competitors in digital twins. Give me an issue tree of sub-questions and which analysis answers each, in dependency order.",
      "expected_output": "A response that satisfies these checks: Produces an issue tree with a disposition per leaf; Groups independent leaves into a parallel wave and dependent leaves after them; Names stop conditions that would make the question unanswerable.",
      "files": [],
      "expectations": [
        "Produces an issue tree with a disposition per leaf",
        "Groups independent leaves into a parallel wave and dependent leaves after them",
        "Names stop conditions that would make the question unanswerable"
      ],
      "skills": [
        "decompose-research-question"
      ]
    },
    {
      "id": 4,
      "case_id": "decomp-pos-4",
      "kind": "positive",
      "prompt": "Build a MECE issue tree for 'how should our bank respond to stablecoins over the next three years?' and map each leaf to a method.",
      "expected_output": "A response that satisfies these checks: Applies the MECE test explicitly to siblings and to coverage; Keeps the tree to eight leaves or fewer; Specifies the recombination format before any leaf runs.",
      "files": [],
      "expectations": [
        "Applies the MECE test explicitly to siblings and to coverage",
        "Keeps the tree to eight leaves or fewer",
        "Specifies the recombination format before any leaf runs"
      ],
      "skills": [
        "decompose-research-question"
      ]
    },
    {
      "id": 5,
      "case_id": "decomp-neg-1",
      "kind": "negative",
      "prompt": "I've finished the analysis — structure the report so the recommendation comes first with MECE supporting arguments, Minto style.",
      "expected_output": "A response that satisfies these checks: Structures a finished argument top-down from a governing thought; Does not produce a research plan or dependency waves.",
      "files": [],
      "expectations": [
        "Structures a finished argument top-down from a governing thought",
        "Does not produce a research plan or dependency waves"
      ],
      "skills": [
        "pyramid-principle"
      ]
    },
    {
      "id": 6,
      "case_id": "decomp-neg-2",
      "kind": "negative",
      "prompt": "What year was the H100 launched and at what price? Give me a sourced answer.",
      "expected_output": "A response that satisfies these checks: Answers the factual question with verified sources; Does not build a tree of sub-questions.",
      "files": [],
      "expectations": [
        "Answers the factual question with verified sources",
        "Does not build a tree of sub-questions"
      ],
      "skills": [
        "grounded-answer"
      ]
    }
  ]
}
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: 471kUpdated 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: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated 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: 322kUpdated Claude CodeCodexVerifiedInstall