guided-research-slice
What It Does
Section titled “What It Does”guided-research-slice is the in-conversation counterpart to research-slice. Where auto-mode dispatches research as an isolated sub-agent with pre-loaded context and depth calibration, the guided version runs directly in the user’s active context window — making the exploration visible and interruptible. The user is present and can steer: pointing out relevant files, clarifying constraints, or stopping to redirect before the research artifact is written.
The prompt instructs the agent to read .gsd/DECISIONS.md and .gsd/REQUIREMENTS.md first, targeting research toward active requirements the slice owns. From there the agent explores relevant code using rg/find for targeted reads or scout for broader unfamiliar subsystems, and checks libraries with resolve_library/get_library_docs for anything not already used in the codebase.
A key framing distinguishes this prompt from raw exploration: you are the scout, not the planner. A planner agent reads the output in a fresh context to decompose the slice into tasks. The research document must be written for that planner — surfacing key files, where the work divides naturally, what to build first, and how to verify. If the research doc is vague, the planner re-explores code the scout already read; if it’s precise, the planner decomposes immediately. This framing keeps the output decision-driving rather than just fact-collecting.
Rather than the elaborate depth-calibration and numbered steps of the auto-mode version, guided-research-slice uses five Strategic Questions to structure the agent’s thinking: what to prove first, what existing patterns to reuse, what boundary contracts matter, what constraints the codebase imposes, and what failure modes should shape slice ordering. The result is a {sliceId}-RESEARCH.md file written to the slice directory, consumed by guided-plan-slice or plan-slice in the next step.
Pipeline Position
Section titled “Pipeline Position”When the user runs /gsd during the planning phase and selects “Research slice first” from the smart entry wizard (shown only when no RESEARCH.md exists yet), the guided flow dispatches guided-research-slice in the current conversation. Once the research artifact is written, the user can run /gsd again to proceed to planning, where guided-plan-slice consumes the RESEARCH.md output.
Variables
Section titled “Variables”| Variable | Description | Required |
|---|---|---|
sliceId | Current slice identifier within the milestone (e.g. S01) | Yes |
sliceTitle | Human-readable title of the slice being researched | Yes |
milestoneId | Current milestone identifier (e.g. M001) | Yes |
skillActivation | Injected skill-loading instruction block; activates skills relevant to the slice context | Yes |
inlinedTemplates | Research output template content inlined directly into the prompt | Yes |
Used By
Section titled “Used By”/gsd— dispatched when the user selects “Research first” during the planning phase of a slice