feat(mystery): build cases that mislead instead of pointing straight at the culprit - #61
Merged
Merged
Conversation
…at the culprit Generated cases were solvable in one move: every clue pointed the same way, no witness ever said anything untrue, and the murder method named the killer's profession outright. The inspector then closed what little gap was left by interpreting each piece of evidence out loud. Generation now demands misdirection: no single clue may identify the killer, the means must be available to at least three suspects, one innocent is framed hard enough to be accused mid-game, and every suspect carries a false_lead — a claim they state as fact but which is wrong — with at least one of them appearing to clear the real killer. The solution gained red_herring and turning_point so the reveal explains the trail the player was meant to follow. Rejected generations now retry with the validation error attached rather than blindly repeating the prompt. The featured villa case was rewritten to the same standard: the poison comes from the victim's own missing heart drops (anyone could take them), a fourth piece of evidence puts the victim alive on the phone describing his own symptoms, and the culprit is reachable only by intersecting the glass, the onset window and the butler's testimony. Viktor now looks guilty for most of the game. The inspector is barred from ranking suspects, confirming theories, or saying what a clue means — he keeps the file, the player runs the case. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ToCUNy2SqwvfTq4a6xfSk1
…ching them check_accusation answered a wrong accusation with the innocent's full not_killer_note, which reasons from the very facts that solve the case — the poisoning window, the phone call, the discarded pen. Guessing therefore paid better than investigating: one bad accusation handed over more of the chain than a careful interrogation. Innocents now carry a separate one-line rebuttal: a single verified fact that blocks the accusation and explains nothing. That is all a wrong accusation returns. The full not_killer_note moves to the debrief after a correct accusation, where check_accusation now also returns cleared_suspects so the inspector can close every thread at once. Generated cases must supply both fields, and the rebuttal is required to avoid any fact appearing in turning_point or evidence_chain — a criterion the tests check against the villa case. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ToCUNy2SqwvfTq4a6xfSk1
The instructions changed in this branch, so a project imported at 1.0 should show a version delta in the template sync view rather than only a field diff. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ToCUNy2SqwvfTq4a6xfSk1
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
2 participants
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.
Generated cases were solvable in one move: every clue pointed the same way,
no witness ever said anything untrue, and the murder method named the
killer's profession outright. The inspector then closed what little gap was
left by interpreting each piece of evidence out loud.
Generation now demands misdirection: no single clue may identify the killer,
the means must be available to at least three suspects, one innocent is
framed hard enough to be accused mid-game, and every suspect carries a
false_lead — a claim they state as fact but which is wrong — with at least
one of them appearing to clear the real killer. The solution gained
red_herring and turning_point so the reveal explains the trail the player
was meant to follow. Rejected generations now retry with the validation
error attached rather than blindly repeating the prompt.
The featured villa case was rewritten to the same standard: the poison comes
from the victim's own missing heart drops (anyone could take them), a fourth
piece of evidence puts the victim alive on the phone describing his own
symptoms, and the culprit is reachable only by intersecting the glass, the
onset window and the butler's testimony. Viktor now looks guilty for most of
the game.
The inspector is barred from ranking suspects, confirming theories, or
saying what a clue means — he keeps the file, the player runs the case.
Co-Authored-By: Claude Opus 5 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_01ToCUNy2SqwvfTq4a6xfSk1