Research AQE Systems
- Read
AGENTS.md, the question, and any user-supplied artifacts completely. - Define the decision the research must support, its scope, and freshness needs.
- Search local evidence broad-to-narrow with
rg --files,rg, code navigation, tests, configuration, documentation, and git history. - For current external claims, inspect primary sources: upstream repositories, releases, commits, issues, pull requests, and official documentation. Record dates, versions, commit identifiers, and links.
- Trace definitions to usages and public boundaries. Map dependencies, data flow, CLI/MCP parity, configuration generation, and tests when relevant.
- Classify each material statement using references/evidence-rules.md. Never repeat a benchmark, capability, or roadmap claim as fact without direct evidence.
- Synthesize only after gathering evidence. State contradictions, uncertainty, missing access, and stale documentation explicitly.
- Return:
- the decision-oriented summary;
- verified findings with source locations;
- inferences and their supporting evidence;
- gaps and risks;
- prioritized recommendations and the smallest useful next validation.
Use parallel agents only when the user requests delegation or repository instructions explicitly permit it. Treat Claude agent definitions as domain references; do not copy or execute their Claude-only hooks or tool syntax.