2 Commits
Author SHA1 Message Date
kathleenandClaude Opus 5.5 d9f3c200b8 Advisor: never advise an option that makes the target impossible while another keeps it possible
CI / test (pull_request) Successful in 57s
On a later attempt the data put the Black near zero on every route, so every option
tied and the bot's tie rule (random among options within 0.2 points) walked away from
the squadmate or went to the valley (DragonKind research runs 90, 91 and 102). The
advisor now also solves the game with no data (every roll possible except through a
closed gate; an encounter bonds its own colour) and marks options after which the
target cannot happen at all as `ruledOut`. The choice is never one of them while
another option keeps the target possible, and recommendOption leaves them out of the
options it gives the bot. For the Black: always investigate, help and climb the tree.

Values are unchanged; the published route tree still matches option for option.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UCzTd31GphpMXDhQqWrCMQ
2026-11-22 02:15:44 +00:00
kathleenandClaude Opus 5.5 77ce54acec Test the bot's call chain in CI; send the experiment plan and screen tree to the bot (#26)
CI / test (pull_request) Successful in 45s
tests/core/bot_contract.test.js copies the bot's AtlasPolicy.advise (DragonKind
src/atlasPolicy.js at ebe134a2) and plays the bot's own recorded screens through it
for every colour, sample and goal; checks the Black is routed through both gates and
up the tree; checks every assignment of the experiment plan is playable; checks the
known answers' shape. Added to npm run test:unit.

push-policy now also sends data/public/experiments.json and tree.json, which
docs/BOT_INTEGRATION.md says the bot reads (Solver.assign, screenPick) but which were
never delivered.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UCzTd31GphpMXDhQqWrCMQ
2026-11-19 21:47:24 +00:00