Synopsis
aigon research-do <ID>Shortcuts
/ard(slash command) ·aigon ard(CLI)
Description
research-do is the main work step in the research lifecycle — the equivalent of feature-do but for investigation rather than implementation. The agent reads the research topic spec, conducts deep investigation (web searches, primary sources, codebase exploration), and writes structured findings with trade-offs, recommendations, and suggested features.
In Drive mode the agent writes directly into the ## Findings section of the main research spec. In Fleet mode (multiple agents) each agent writes only to its own findings file (research-{ID}-{agent}-findings.md), leaving all other files untouched.
Research agents work in the main repository on the current branch — unlike feature agents, they do not need branch isolation because they only write findings files, not source code.
What it does
- Reads the research topic spec from
docs/specs/research-topics/03-in-progress/research-{ID}-*.md - Signals
implementingstatus to the dashboard viaaigon agent-status implementing - Conducts research: web searches, documentation reads, codebase exploration
- Writes findings covering consensus views, divergent approaches, trade-offs, and source citations
- Fills in the
## Suggested Featurestable with feature names, descriptions, priorities, and dependencies - In Fleet mode, writes only to the per-agent findings file; in Drive mode, writes to the main spec
Examples
# Drive mode — agent investigates and writes findings directly to the spec
aigon research-do 5
# Fleet mode — each agent runs this command in its own session
/aigon:research-do 5Key constraints
- Do not modify source code (
.js,.ts,.py, etc.) — research is read + write of markdown only - Do not modify other agents’ findings files in Fleet mode
- Do not create new branches — stay on the current branch
Related
research-start— creates the workspace and starts agentsresearch-open— opens Fleet agent sessions side-by-sideresearch-submit— Fleet agents call this when findings are completeresearch-eval— synthesizes findings from all agents after research-do