research-workflow

Plugin research-workflow from marketplace agent-workflows in destenson/agent-workflows. Commands read from the repo's manifest — 1 month ago.

Skills (13)

• decision-memo

  Draft a decision memo — ship, iterate, or kill. Use after a results memo is finalized. Closes the loop from proposal to outcome and, if the decision is iterate or kill, spawns the next node in the tree.

• experiment-design

  Draft an experiment plan — the methods protocol that connects a proposal's hypothesis to the run that will test it. Use after a research proposal is approved and before any experiments are run. Produces the document that run logs and eval reports derive from.

• gap-analysis

  Conduct a gap analysis and generate hypotheses — identify what the prior art does not cover and formulate testable hypotheses about how to address it. Use after a literature scan, before writing a research proposal.

• generate-feature-proposal

  Draft a feature proposal — "should we add this to a thing that exists?" Use when the ask is scoped to an existing product and the decision is user/maintenance cost vs. value.

• generate-product-proposal

  Draft a product proposal — "should we build a new thing?" Use when the ask is a greenfield product or standalone system, requiring market/user framing and a demand bet.

• generate-project-proposal

  Draft a project proposal — "should we resource this effort?" Use when the primary question is about cost, timeline, and people rather than what to build or whether a hypothesis is true.

• generate-research-proposal

  Draft a research proposal — "is this hypothesis worth testing?" Use when the goal is a learning, not a shipped artifact. The output is a decision gate, not a funding document.

• literature-scan

  Conduct a literature review and prior-art scan — survey what has already been tried, internally and externally, before committing to a hypothesis or proposal. Use upstream of a research or product proposal. Produces a prior_art_scan node in the document tree.

• negative-results

  Document a negative result — a falsified hypothesis, a failed approach, or an inconclusive experiment. Use after a results memo or when an experiment is stopped early. The most underrecorded document type in the tree, and often the highest reuse value.

• preregistration

  Write a preregistration — a public or internal commitment to an analysis plan, made before data is collected or results are seen. Use for empirical ML work to prevent post-hoc rationalization of results. Produced alongside or immediately after the experiment plan.

• problem-statement

  Draft a problem statement — the bounded, falsifiable description of what is wrong or unknown and who it affects. Use before writing a proposal of any type. A sharp problem statement is the prerequisite for a useful hypothesis.

• research-init

  Scaffold the research workflow in this project — a research/ directory with the common starter artifacts (abstract, proposal, experiments, results), customized to the project by autodiscovery and a short interview. Use when the user asks to initialize, scaffold, or set up the project's research work

• results-memo

  Draft a results memo — the eval report that synthesizes run logs and measurements into findings. Use after experiments are complete. Produces the document that justifies the decision memo.

Hooks (3)

• SessionStart

  ${CLAUDE_PLUGIN_ROOT}/scripts/session-start.sh

• UserPromptSubmit

  ${CLAUDE_PLUGIN_ROOT}/scripts/standing-rules.sh

• Stop

  ${CLAUDE_PLUGIN_ROOT}/scripts/distillation-gate.sh

Similar plugins