caee558e79
Add /ultraresearch-local for structured research combining local codebase analysis with external knowledge via parallel agent swarms. Produces research briefs with triangulation, confidence ratings, and source quality assessment. New command: /ultraresearch-local with modes --quick, --local, --external, --fg. New agents: research-orchestrator (opus), docs-researcher, community-researcher, security-researcher, contrarian-researcher, gemini-bridge (all sonnet). New template: research-brief-template.md. Integration: --research flag in /ultraplan-local accepts pre-built research briefs (up to 3), enriches the interview and exploration phases. Planning orchestrator cross-references brief findings during synthesis. Design principle: Context Engineering — right information to right agent at right time. Research briefs are structured artifacts in the pipeline: ultraresearch → brief → ultraplan --research → plan → ultraexecute. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1.7 KiB
1.7 KiB
Config-Audit UX Rules
These rules apply to ALL config-audit commands. The goal is a professional, human-friendly experience.
Output Rules
- NEVER show raw JSON, stderr output, or scanner progress lines to the user
- ALL scanner Bash commands MUST use
--output-file <path> 2>/dev/null - Check exit code via
; echo $?— codes 0, 1, 2 are normal (PASS/WARNING/FAIL). Only 3 is a real error - Read output files with the Read tool, extract key metrics, and present formatted results
- NEVER let the user see tool call output that looks like diagnostic logs or stack traces
Narration Rules
- Before each major step, tell the user what's happening in plain language
- After scanners complete, briefly say what was found before showing details
- When spawning agents, tell the user what the agent does and approximate wait time
- If something takes more than a few seconds, set expectations: "This takes about 30 seconds..."
Formatting Rules
- Use markdown tables for structured data (area breakdowns, finding lists)
- Add one-sentence plain-language context for grades and scores — don't assume the user knows what "Level 4 Governed" means
- Separate test-fixture/example findings from real findings when showing counts
- End every command with context-sensitive next steps — explain what each command does, not just its name
- Adapt tone to results: A/B grades get encouraging context, D/F grades get empathetic, actionable guidance
Command Format
- Always use space-separated format in suggestions:
/config-audit plan(NOT/config-audit:plan) - Never reference commands that don't exist
- When suggesting next steps, explain WHY the user might want each option