add_assumption
ChatGPTRecord something that must be true for the idea to work, how risky it is if false, and how it will be tested. Pair it with a stop rule on the idea (update_idea stop_rule).
add_reference
ChatGPTAttach a link to a chat, note, external page or prototype.
append_note
ChatGPTAppend a dated entry to an idea's notebook (research, a decision, a conversation summary).
archive_idea
ChatGPTFile an idea away with a reason. It keeps its score, notes and history and can be restored. For merged, give merged_into_id.
create_idea
ChatGPTPin a new idea to the board. Raw text is enough. Vibestash then fills in whatever sections are still empty (pitch, problem, audience, competitors) and scores it in the background; fields you pass are kept as they are. Set enrich=false to skip filling in and scoring.
delete_idea
ChatGPTIrreversible. Prefer archive_idea. Requires a token with the delete scope and the exact idea title as confirm.
delete_note
ChatGPTRemove a wrong or outdated notebook entry. The history keeps a line saying it was deleted.
get_build_brief
ChatGPTEverything needed to start building an idea in one call: the latest versioned build brief (Markdown + JSON), all sections, assumptions, evidence, notebook, references and recent history. If the idea has no brief yet, the first call writes version 1; later calls return it unchanged.
get_idea
ChatGPTAn idea with its sections, next step, stop rule, assumptions, evidence log, notebook, references and links.
get_profile
ChatGPTHow the builder builds and markets (0–4 each), areas they know, hours a week, budget and risk appetite. Personal fit and time to live are judged against it.
get_research
ChatGPTThe latest Vibestash deep research report for an idea, if one was run from the web app: its progress, or the summary, path to market and first move when done (the full report is in the idea's notebook). For new research in the chat use research_plan.
get_score_history
ChatGPTEvery score of an idea, oldest first: which factors moved, from what to what, why, and what caused it (a rescore, new evidence, an edit by you or an assistant).
link_ideas
ChatGPTMark two ideas as related.
list_ideas
ChatGPTList ideas on the board or in the archive, optionally filtered.
list_notes
ChatGPTAn idea's notebook entries, oldest first, with their ids for update_note and delete_note.
log_evidence
ChatGPTRecord real evidence of demand: interviews, survey responses, sign-ups, waitlist sign-ups, pilots, letters of intent, pre-orders, payments. Set supports=false for evidence against (people said no, a pilot was cancelled). Demand is then rated from the evidence log, strongest commitments counting most, and the idea is rescored unless demand is locked. Only log what actually happened.
move_idea
ChatGPTMove an idea along the lifecycle (spark, exploring, validated, building, shipped).
remove_assumption
ChatGPTDelete an assumption that no longer applies. The history keeps a line saying so.
remove_evidence
ChatGPTDelete a wrong evidence entry; demand is rated again from what is left.
report_progress
ChatGPTReport how many tasks of a Building idea are done ("14 of 22"), or send the TASKS.md contents to derive it.
research_plan
ChatGPTCall this when the user asks to research an idea's market. Returns what to look up for this idea (demand, competitors and their prices, the path to market) and the rules for sources. Do the searches with your own web tools, then save the findings with save_assistant_research. Changes nothing.
restore_idea
ChatGPTRestore an archived idea to the stage it left.
save_assistant_research
ChatGPTSave market research you did with your own web search (follow research_plan first). Replaces the market summary, and the competitors if you send them; demand signals, channels, the first move and notes go into a notebook entry labelled Assistant research. Every source must be a real page you read.
score_idea
ChatGPTSet any of the four factors 0–10 (demand; competition, 10 = open field; effort, 10 = easy; personal fit) with reasons, and/or set ai=true to have Vibestash rescore the factors you did not set. lock keeps factors fixed from now on, so "keep effort at 4, rescore the rest" is {effort: 4, lock: ["effort"], ai: true}. With logged evidence, demand comes from the evidence. Score = 35% demand + 25% competition + 15% effort + 25% personal fit. A first score needs all four factors or ai=true.
search_ideas
ChatGPTFull-text search across live and archived ideas, including notebook entries.
set_profile
ChatGPTReplace the builder profile. engineering (how they build): 0 = by chatting with AI builders like Lovable or Bolt (a real way to ship; many ideas need nothing more), 1 = can follow what the AI proposes, 2 = can set up a database and sign-in with guidance, 3 = makes architecture and security calls, 4 = has run production systems. It is judged against each idea's complexity, never on its own. go_to_market: 0 = not yet, 2 = can run a launch and talk to customers, 4 = it is their day job. hours_per_week sets how fast an idea can go live. Only record what the user told you; rescore ideas afterwards (score_idea ai=true).
sync_progress
ChatGPTRead progress from the idea's linked public GitHub repo: TASKS.md checkboxes, else the open milestone due soonest, else all issues. Synced ideas stay current every six hours while Building.
update_assumption
ChatGPTChange an assumption, its test, or record what happened: status testing, confirmed or disproved, with the result.
update_idea
ChatGPTUpdate fields or sections of an idea. Only the fields you pass change.
update_note
ChatGPTChange the title or body of a notebook entry. The history records that it was edited.