Structures design interviews at one fixed altitude to agree system decomposition and scope.
npx skills add https://github.com/vinzenz/workflows --skill survey-vinzenz
Please help me install this Agent Skill. Skill: survey Source: https://github.com/vinzenz/workflows/tree/main/.agents/skills/survey Command: npx skills add https://github.com/vinzenz/workflows --skill survey-vinzenz
High-intent search queries and answers about installing and using this skill.
Run a survey at the A1 Parts altitude: draft a strawman decomposition of every part, ask 4-6 recommendation-first questions per round, and read back a numbered standing of agreed, open, and parked items until the level's exit test passes.
Survey works breadth-first at one fixed altitude, covering the whole level before sharpening anything, and outputs the level's shape plus wanted/not-wanted lists. Grilling works depth-first, descending wherever the interesting question is, and outputs decisions and ADRs.
Skip it when the question has one obvious answer, when the level cannot be settled by talking and needs a prototype, when the effort spans several altitudes (use pathfinder), or when the shape is agreed and only one deep decision remains (use grilling).
State the altitude and what is off-limits before the first question. When a lower-altitude question arrives, name it, park it with its altitude label, and return the conversation upward in the same breath.
No. A survey writes only CONTEXT.md glossaries, ADRs, spec text, and its own scratch notes. If implementation feels like the next step, the survey is done and the work should be handed off.
forcedotcomStress-tests plans against domain glossaries and architecture decision records through structured questioning.
jstoobzGenerate TLDR.md, CONTEXT_FOR_NEXT_SESSION.md, and PROMPT_LAB.md by merging outputs with existing files.
onesyntaxAnalyze software component design using cohesion, coupling, and dependency principles.
zhengfranInterview users one question at a time to clarify plans and resolve decision branches.
