synero — CLI для получения экспертных оценок и анализа

--- name: synero description: Ask Synero’s AI Council questions from the terminal and get one synthesized answer from four contrasting AI advisors. Use when a user wants a higher-confidence take on strategy, research, architecture, hiring, positioning, or other judgment-heavy questions where multiple perspectives beat a single-model reply. metadata: { "openclaw": { "requires": { "bins": ["python3"] }, "env": ["SYNERO_API_KEY"], "optionalEnv": [ "SYNERO_API_URL", "SYNERO_TIMEOUT", "SYNERO_MODEL_ARCHITECT", "SYNERO_MODEL_PHILOSOPHER", "SYNERO_MODEL_EXPLORER", "SYNERO_MODEL_MAVERICK", "SYNERO_MODEL_SYNTHESIZER" ] } } --- # Synero Skill Use this skill to query Synero from the terminal when one AI answer is not enough. Synero is positioned as a collective AI intelligence product: one prompt goes to four advisors with different reasoning styles, then Synero synthesizes them into a single response. This is most useful for ambiguous or high-leverage questions where you want disagreement, tradeoffs, blind-spot checking, and a clearer final recommendation. ## What this skill helps with Use it for prompts like: - product or roadmap decisions with meaningful tradeoffs - technical architecture reviews and go/no-go questions - research questions where cross-checking perspectives improves trust - hiring, org, or leadership choices with second-order effects - messaging, positioning, or content strategy that benefits from competing angles Prefer simpler tools for basic factual lookups or one-shot chores. This skill shines when the question is judgment-heavy. ## What it does - Sends a prompt to Synero’s council endpoint - Supports 4 advisor roles: `architect`, `philosopher`, `explorer`, `maverick` - Returns either: - a synthesized final answer for normal use, or - raw SSE stream output for debugging and live visibility - Supports optional thread continuity and per-slot model overrides ## Prerequisites Get your API key from `https://synero.ai`, then export it before running the script: ```bash export SYNERO_API_KEY="sk_live_..." ``` If you are not sure where to get the key, sign in at `https://synero.ai` and use the API/settings area there. Optional environment variables: ```bash export SYNERO_API_URL="https://synero.ai/api/query" export SYNERO_TIMEOUT="120" export SYNERO_MODEL_ARCHITECT="gpt-5.2" export SYNERO_MODEL_PHILOSOPHER="claude-opus-4-6" export SYNERO_MODEL_EXPLORER="gemini-3.1-pro-preview" export SYNERO_MODEL_MAVERICK="grok-4" export SYNERO_MODEL_SYNTHESIZER="gpt-4.1" ``` ## Quick command ```bash python3 ~/.openclaw/skills/synero/scripts/synero-council.py "Should we ship this feature in the next 30 days?" ``` That command uses `SYNERO_API_KEY` from your environment and sends the request to Synero at `https://synero.ai/api/query` unless you override `SYNERO_API_URL`. ## Quiet final-output mode Use `--quiet` when you want only the synthesized answer with no extra status lines: ```bash python3 ~/.openclaw/skills/synero/scripts/synero-council.py --quiet "Evaluate this architecture plan and recommend a prototype path." ``` ## Streaming and debugging mode Use `--raw` when you want the raw event stream for troubleshooting or to inspect council behavior live: ```bash python3 ~/.openclaw/skills/synero/scripts/synero-council.py --raw "What are the strongest arguments for and against this pricing change?" ``` ## Advanced configuration Use model overrides when you want to pin specific advisors or a specific synthesizer: ```bash python3 ~/.openclaw/skills/synero/scripts/synero-council.py --thread-id "your-thread-id" --advisor-model architect=gpt-5.2 --advisor-model philosopher=claude-opus-4-6 --advisor-model explorer=gemini-3.1-pro-preview --advisor-model maverick=grok-4 --synthesizer-model gpt-4.1 "Your question" ``` ## Output behavior Default mode prints: - HTTP status line - final synthesized answer `--quiet` prints: - final synthesized answer only `--raw` prints: - raw SSE events from the API ## Prompting guidance Write prompts that invite useful disagreement, not generic brainstorming. - Ask for tradeoffs, risks, assumptions, and a recommendation - Include operating constraints such as timeline, budget, team size, and risk tolerance - Use `--thread-id` when continuing the same topic across multiple rounds - Use `--quiet` when another tool or script needs clean final text only - Use `--raw` when debugging streaming behavior or inspecting advisor outputs For reusable prompt templates, read: - `references/prompt-patterns.md` ## Error handling - Missing key → exits with clear guidance to set `SYNERO_API_KEY` - HTTP failure → prints status and response body - Network failure → prints a clear network error - Empty synthesis → exits non-zero instead of pretending things are fine