Overview
/silver:compare is a thin orchestrator route into /silver:deep-research with research_type=solution-compare. It compares N ≥ 2 named solutions you supply — no market shortlist step. The deep-research engine runs the full AF-DECIDE nested workflow, produces SCRs and a weighted matrix, and packages a serverless SPA report.html you open directly in the browser (no HTTP server).
Output lives under research/<date>-<slug>/ (for example research/2026-07-09-compare-backstage-port-cortex/). Does not invoke MultAI or multi-AI orchestration.
Specialized path into WF-SILVER-DEEP-RESEARCH / FS-SILVER_DEEP_RESEARCH — not a separate multi-AI workflow. See /silver:deep-research for the full research engine, modes, and handoff options.
/silver:deep-research with research_type=solution-landscape when you need a market survey and top-5 shortlist. Use /silver:compare when you already know which solutions to evaluate.
When to Use
Entry trigger signals for /silver:compare:
- You have two or more named solutions to evaluate (tools, platforms, vendors, frameworks)
- "compare Backstage vs Port vs Cortex" / "evaluate these three options"
- Weighted decision matrix with evidence-backed SCRs per solution
- You do not need a market landscape shortlist — the user list is the research set
Reject if fewer than two solutions. Soft-warn if more than twelve (continue unless user confirms).
Catalog Composition
| Atomic flow | Command | Outcome |
|---|---|---|
AF-CLARIFY | /silver:clarify | Need-profile interview; need_profile.json with interview_complete: true |
AF-DECIDE | /silver:compare → /silver:deep-research | SCR per solution, weighted matrix, report.html |
AF-DOCUMENT | /silver:ensure-docs | Durable artifacts under research/<date>-<slug>/ |
AF-VALIDATE | /silver:validate | validate_compare.py + validate_spa_report.py gates |
Workflow Steps
1. Parse named solutions
Parse N ≥ 2 solution names from the invocation. Optional URLs or repos in parentheses. Write solutions_requested.json with the named list and run_manifest.json with research_type: solution-compare.
2. Need-profile interview
Mandatory before retrieval (phase_gate.py blocks DR-RETRIEVE until complete). Runs via clarify discipline using reference/need-profile-interview.md. Captures must-haves, license preference, constraints, and success criteria in need_profile.json.
With --auto (only when user explicitly passes it): record assumed defaults with "auto_assumed": true. Never silently invent must-haves.
3. Deep-research engine (DR-* phases)
Delegates all DR-* phases to skills/silver-deep-research/SKILL.md. Default mode for solution types is deep. Skips market shortlist — user list is the research set.
4. SCR extraction per solution
Produces a Solution Capability Report (SCR) per named solution under solutions/<slug>/scr.md plus features.json. Evidence spans, capability scoring, and citation lineage roll into the comparison.
5. Weighted matrix and SPA report
compare_solutions.py builds comparison/comparison.json and comparison/comparison-matrix.md. generate_report_spa.py emits serverless report.html — open in Cursor browser and system browser (open report.html). No http.server or live server required.
Output Artifacts
Canonical output directory:
Research lineage is preserved for downstream /silver:feature, /silver:ui, /silver:devops, or /silver:plan handoff — same as deep-research handoff.
Validation
Standard deep-research validation (validate_report.py, citation and claim-support checks, phase_gate.py) also applies before exit.
Example Invocation
Silver Bullet parses three named solutions, runs the need-profile interview, creates research/2026-07-09-compare-backstage-port-cortex/, produces SCRs and a weighted matrix, validates, and opens report.html. Hand off to implementation workflows when the decision is ready.