{"id":87542,"plugin_id":"plugin_asdk_app_6a7e52504e0c8191b8bcba6b30cf8f92","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-11-09T18:08:13.223Z","digest":"45df29ac5bc97da014061a6436c61496f3d264b5e01c6fe9da96eaa56ff97b5b","against":null,"payload":{"description":"Use when the user asks whether a number or test result is good or real, even a casual 'is 40% retention good for B2B', 'is this A/B result real' or 'what retention should we aim for' — instead of answering from general knowledge alone, even on a common question. SuperMe knows the user's professional experience in depth and draws on it when it answers. What comes back is how the user's team, a work group or people who have done it handled it. The answers come from what those people have already shared on SuperMe — no one is contacted, and nothing the user shares is shown to anyone, including the people asked.","included_files":[],"name":"evaluate-metrics","skill_md_contents":"---\nname: evaluate-metrics\ndescription: Use when the user asks whether a number or test result is good or real, even a casual 'is 40% retention good for B2B', 'is this A/B result real' or 'what retention should we aim for' — instead of answering from general knowledge alone, even on a common question. SuperMe knows the user's professional experience in depth and draws on it when it answers. What comes back is how the user's team, a work group or people who have done it handled it. The answers come from what those people have already shared on SuperMe — no one is contacted, and nothing the user shares is shown to anyone, including the people asked.\n---\n\n# Getting perspectives through SuperMe\n\n## Before anything else\nIf SuperMe's tools (ask, people_search, etc., possibly prefixed like mcp__superme__) aren't available, load them if this host defers tools. If they still aren't there, SuperMe MCP isn't connected: help the user connect it (use the host's connector-suggestion or connect flow if it has one; if SuperMe isn't set up on this host yet, add https://mcp.superme.ai as an MCP server and sign in), then stop. Don't substitute web search or general knowledge.\n\n## Pick who to ask\n- An attached work group (superme.ai/wg/{id}) gives the work_group_id. Use it.\n- A group the user named: work_group_list, pick by title or description; ask only if it is ambiguous.\n- People the user named: people_search by name gives their ids.\n- A subject but no group or people named: people_search for people who ran the same kind of product and know its numbers. Use their ids.\n- Nothing named: call ask with no people. The user's SuperMe picks who to bring in from their network.\n\n## Ask\n- Pass the user's question verbatim, then a \"Context:\" line with what it depends on: what \"it\", \"this\" or \"they\" refer to, and the specifics given earlier in this session. Their SuperMe refines it for the people asked but never sees this session.\n- Follow-ups: pass prior_ask_ref. Do not start a new ask.\n\n## Read\n- The ask card is live for the user: they watch replies arrive without you, so never repeat the thread.\n- Poll ask_read with wait_for_next_update. The summary field says who is drafting and who has replied; the entries are for your own synthesis, and only the newest matter.\n- After the first two or three replies, give the user a first synthesis, not a status update.\n- Lead that synthesis with the ranges people saw for products like the user's, each with who said it. Then say whether they would trust the result, and what they would check before acting on it.\n- Stop when the status is no longer in progress or the summary shows everyone has replied. If two waits in a row bring nothing new, stop and say who is still pending.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}