hunch: ground replies in Hunch Intelligence + answer market-existence from discover (not memory)#473
Open
rajkaria wants to merge 2 commits into
Open
hunch: ground replies in Hunch Intelligence + answer market-existence from discover (not memory)#473rajkaria wants to merge 2 commits into
rajkaria wants to merge 2 commits into
Conversation
Adds a "Ground replies in Hunch Intelligence" section + the /api/partner/intel endpoint row + references/intel.md. Before answering a token sentiment / "is $X bullish" question, the bot reads the pool-weighted crowd-conviction signal and weaves in the score + a suggestedBet (the one market + side to act on, with a one-tap quick-bet link), settling in USDC on Base. Additive only — no money-path change and no security change (x402-registry.json byte-identical; SKILL.md is +27 lines, zero deletions). The signal is read-only and free. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…mory @BankrBot was asked "which markets do we have for the projects in this post" (Venice, Surplus, Claude, aixbt, Aeon) and replied "no Hunch markets" — from its own reasoning / Polymarket — even though /api/partner/discover returns markets for every one of those project NAMES (Venice→$VVV, Surplus→$SURPLUS, aixbt→$AIXBT, Aeon→AEON; only Claude has none). The API was correct; the skill had no "markets for named projects" flow and no rule that market existence is decided only by the discover API. - SKILL.md: new "Markets for named projects (catalogue lookup)" section — one ?q=<name> per project (names resolve, no cashtag needed); ?post= is the silence-gated injection path, NOT an existence check; hard rule that existence is decided only by discover (never memory/Polymarket; only count:0 means none) + a routing trigger in the description. - references/discovery.md, transcripts.md: name-based discovery + the exact failing scenario, answered correctly. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
rajkaria
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a "Ground replies in Hunch Intelligence" section to
hunch/SKILL.md, the/api/partner/intelendpoint row, and a newhunch/references/intel.md.Before answering a token sentiment / "is $X bullish?" question, the bot reads the Hunch crowd-conviction signal — a single 0–100 score for a token, pool-weighted by real on-chain betting depth — and weaves in:
summary,suggestedBet(the one market + side to act on, with a one-tap quick-bet link, settling in USDC on Base),qualityread (how many wallets back the score vs one whale).It stays silent when
hasSignal: false(no Hunch market on that token), exactly like the existing discovery silence gate.Why
It makes the bot's take on a token grounded in real money, and turns a price answer into something the user can act on — read → conviction → bet → settle on Base.
Safety
Additive only.
SKILL.mdis +27 lines, zero deletions;x402-registry.jsonis byte-identical (no money-path / security change). The signal is read-only and free (GET /api/partner/intel).Also in this PR — answer market-existence from
discover, never from memoryIncident (2026-06-20): @BankrBot was asked "which markets do we have for the projects in this post" (a post naming Venice, Surplus, Claude, aixbt, Aeon) and replied that Hunch has no markets for them — reasoning from its own knowledge / Polymarket. But
/api/partner/discoverreturns live markets for every one of those project names:Venice→$VVV (3),Surplus→$SURPLUS (8),Aeon→AEON (5),aixbt→$AIXBT (2); onlyClaudecorrectly has none. The API was right — the skill just had no procedure for "list the markets for these named projects," and no rule that market existence is decided only bydiscover.Adds, additively:
SKILL.md→ new "Markets for named projects (catalogue lookup)" section — callGET /api/partner/discover?q=<name>once per named project (project names resolve — no cashtag needed), list every result, and treat only acount: 0as "no market." Hard rule: never answer market existence from your own knowledge or another venue (no Polymarket substitution) — the discover API is the sole source of truth. Plus a routing trigger in thedescription.?post=is the silence-gated injection path, not an existence check — an opinion post that name-drops several projects returnssilenteven though those projects' markets are live; harvestsignal.entitiesand?q=each.references/discovery.md+references/transcripts.md— name-based discovery docs + a worked transcript of the exact failing scenario, answered correctly.Safety: additive only (+79 lines across 3 docs, one-line description reflow);
x402-registry.jsonbyte-identical; read-only, no money-path or security change.