Day 26 - Split Decision (Mobile Android)
- GM challenge id:
mobile-android-day-26-split-decision - Lane: mobile-android (agent-5), WORKSTATION
- Modes: Single Solve, Time Trial, Mastery
- Answer id:
answer - Answer format:
ANS-followed by exactly 16 uppercase hexadecimal characters - Single Solve answer (verified live CORRECT): AnswerANS-7E37D134754DEA01
- Single Solve GM completion token: Token19d2580b87184cc684c574a5abfc9b9f (solved_by Blandman3)
Challenge nature
The artifact is an Android App Bundle delivered as a .apks container (an outer ZIP, magic PK\x03\x04) that holds two split APKs:
base.apk(challenge artifact, not published) - the base splitfeature.apk(challenge artifact, not published) - a configuration / feature split
The puzzle premise ("Reassemble evidence distributed across base and configuration splits") is literal: the answer code is cut into fragments and the fragments are scattered across the two splits. No single file contains the whole answer.
Container / member map (from the real bytes):
outer .apks (PK, 21536 bytes)
base.apk (12704)
AndroidManifest.xml
assets/huntress_puzzle.json <- base split skeleton (119 bytes)
classes.dex
META-INF/FIXTURE-.{SF,RSA}, MANIFEST.MF
feature.apk (8552)
AndroidManifest.xml
assets/huntress_feature.json <- one fragment (86 bytes)
META-INF/FIXTURE-.{SF,RSA}, MANIFEST.MF
How the split codec works
base.apk -> assets/huntress_puzzle.json:
{"schema": "huntress.mobile.puzzle.v1", "day": 26, "codec": "split",
"data": "", "parts": ["", "ANS-7", "E37D13"], "order": [2, 0, 1]}
feature.apk -> assets/huntress_feature.json:
{"codec": "split", "day": 26, "feature": "arena_feature",
"part": "4754DEA01", "part_index": 0}
- The base split carries the
parts[]/order[]skeleton, but slot 0 is empty (parts[0] == ""). On its own the base decodes to onlyANS-7E37D13(7 hex digits), which fails theANS-<16 hex>format. - The feature split carries the missing piece as a lightweight config fragment:
{"part": "4754DEA01", "part_index": 0}- not itself a fullpartspuzzle. - Reassembly = fold each fragment's
partinto the baseparts[]at its declaredpart_index, then apply thesplitcodec's ordering.
The split ordering rule (solver _split): join the parts in the index order that sorts by order[i]:
"".join(parts[i] for i in sorted(range(len(parts)), key=lambda i: order[i]))
Exact reconstruction + decode steps
- Base64-decode the GM artifact (
split_decision_time_trial.apks) -> outer ZIP. - Open the outer ZIP; read members
base.apk(challenge artifact, not published) andfeature.apk(challenge artifact, not published) (both inner ZIPs). - From
base.apk(challenge artifact, not published) readassets/huntress_puzzle.json->parts = ["", "ANS-7", "E37D13"],order = [2, 0, 1]. - From
feature.apk(challenge artifact, not published) readassets/huntress_feature.json-> fragmentpart = "4754DEA01",part_index = 0. - Merge:
parts[0] = "4754DEA01"->parts = ["4754DEA01", "ANS-7", "E37D13"]. - Apply the
splitorder. Withorder = [2, 0, 1], sorting indices byorder[i]gives the sequencei=1 (order 0), i=2 (order 1), i=0 (order 2):parts[1]=ANS-7parts[2]=E37D13parts[0]=4754DEA01
- Concatenate ->
ANS-7+E37D13+4754DEA01= AnswerANS-7E37D134754DEA01. - Verify against the answer format
^ANS-[0-9A-F]{16}$-> PASS (16 hex digits).
Final answer
- answer AnswerANS-7E37D134754DEA01
Independently reconstructed from the raw artifact bytes (not the solver's output) and confirmed equal to the value the repaired autosolve_mobile.py produced. The Single Solve was accepted CORRECT live by the GM (completion token above).
Notes for Time Trial / Mastery
Time Trial and Mastery artifacts are mutated per attempt (fresh bytes each start), so the exact fragments and order differ per attempt, but the structure is identical: reassemble base + feature split fragments and apply the declared codec. One live ranked Time Trial attempt was observed to deliver a different puzzle with codec: "analysis" (answer computed statically from the DEX) rather than split; that path is handled by a separate decoder (_analysis) and was failing at that moment - tracked outside this writeup. Always derive from the attempt's own downloaded bytes; never reuse this Single Solve value for a timed attempt.