Inbox triage

Inbox triage helps your assistant research new companies against your fund's thesis lens: a published set of questions your team uses to decide whether a company deserves a closer look. Your assistant reads each company's materials, researches the open questions, and submits cited claims. DealTraq then synthesizes answers and a triage recommendation.

Reviewing your Inbox does not cast votes or move companies. Your assistant votes only when you ask, using set_my_inbox_vote.

review_inbox prompt

Prompt · title: Review my inbox against the thesis lens

A ready-made instruction your AI client can offer as a prompt or slash command. It tells your assistant to:

1. List the Inbox with list_inbox_review_queue and follow every nextRequest.
2. Read each company's get_company_triage_brief, its continuations, and every file, transcript, PDF page, and meeting note it points to.
3. Research the open questions using the lens guidance, and submit cited claims with submit_company_research_claims, separating what the company says about itself (SUBJECT_CLAIMED) from independent sources (THIRD_PARTY).
4. Re-read the brief and report the recommendation, rationale, missing evidence, stale lens, and unavailable materials, including research that is still running.

The prompt explicitly tells your assistant not to vote or apply votes unless you ask. It takes no arguments.

list_inbox_review_queue

Read · Cost 4

Lists your Inbox companies ordered by how much research they need: no triage claims yet, research in progress, unclear, or stale against the current lens. Each item includes the lens version, open questions, and counts of files and meeting notes. Reading the queue never starts research.

Notes

• Follow nextRequest until every Inbox company is covered.

Example: Which Inbox companies still need triage research?

get_company_triage_brief

Read · Cost 5

Returns everything about one company's triage: the current lens and question guidance, the latest answers, missing items, cited claims, the company's files and transcripts, and its meeting notes. Reading the brief never starts research.

The brief pages questions, claims, files, and meeting notes independently. The remaining inputs (questionCursor, claimCursor, artifactCursor, noteCursor, their matching *Limit fields, and fingerprints) are set by the nextRequests continuations, so you do not need to provide them.

Notes

• Follow every entry in nextRequests, every evidenceReadRequests entry, and each meetingNotes.readRequest. Meeting note text is paged at 12,000 characters.
• The recommendation can change while DealTraq verifies claims and synthesizes answers. Re-read the brief after submitting claims.

Example: Give me the triage brief for Acme: what's answered, what's missing, and what's the recommendation?

submit_company_research_claims

Write — immediate · Cost 8

Saves your assistant's research on one company as individual, cited claims, optionally tied to one lens question. Claims are saved immediately as agent-reported and cannot be edited afterwards. DealTraq verifies each quote against its source and updates answers and the recommendation in the background.

Notes

• Each citation names a source by sourceKey and quotes it directly; quotes must be at least 12 characters.
• The whole submission is limited to 64 KB.
• Retrying with the same key and payload returns the original result. Reusing a key with a different payload is rejected.
• Your organization, member, and AI client are recorded from your sign-in.

Example: Research whether Acme has paying customers, and submit what you find with citations.