Tactical LogHuntress CTF 2026 // team archive

Days / Day 21 / Warmups / Under Pressure

DAY 21//LANE: WARMUPS//STATUS: CORRECT x3

Under Pressure

Warmups SSTTMM Modes reported CORRECT in this write-up

Under Pressure (Day 21, warmups)

GM id: warmups-day-21. Name: 👶 Under Pressure. Statement SHA-1: 5c951cd4c65bf33ed5b23eb43f89b33f760ed9c3. Mastery blob: 103 sets, each with 21 numbered lines; the first set*.message sits at line 1, each other field is a positional encoding of the one plaintext.

Result

All three modes CORRECT, accepted by the GM (solved_by Blandman3):

ModeResultAccepted answerRanked
SingleCORRECTToken6eb658e74fc8cf7b8a09bddaab83c084no
Time TrialCORRECT (1.3 s)Token79f267098d8b51385207b41801250d71yes
MasteryCORRECT (3.3 s, retry)Token3a7e980f775c6456495564056cc89ce3no

Method

The warmups adapter (challenges/_lib/autosolve_warmups.py) walks each set's lines against a fixed positional decoder ladder in LADDER_NAMES. Line 1 is the plaintext; lines 2..n are its encoded forms, in a known order. Reading any set and matching the position against the ladder recovers the one message.

Single and Time Trial were solved in the first autosolve pass. The first ranked Mastery attempt (3d9848c7...) ran against a blob that reached line 21, and the ladder in LADDER_NAMES stopped at line 18, so lines 19–21 fell through, the pass could not resolve the three sets whose message lived at those positions, and the timed window expired at 13:04:38Z unsolved.

The decoder gap

Lines 19, 20, 21 of each set were, respectively:

  • utf-16-le of the plaintext, encoded as a lowercase hex string carrying its BOM
  • utf-16-be of the plaintext, encoded as a lowercase hex string carrying its BOM
  • gzip of the plaintext bytes, then base64 (standard, with padding)

The three decoders were added to LADDER_NAMES (positions 19..21) and to the dispatch table in autosolve_warmups.py:

  • utf16lehex: bytes.fromhex(s).decode("utf-16-le") and discard the BOM
  • utf16behex: bytes.fromhex(s).decode("utf-16-be") and discard the BOM
  • gzipb64: gzip.decompress(base64.b64decode(s)).decode("utf-8")

Offline, the fix read all 103 sets of the expired blob without a miss.

Retry route fix

Autosolve's post-ranked-failure path (_fallback_with_live_status in challenges/_lib/autosolve.py) refused to send a continuation when the GM status was expired without a next_op field — the state this attempt was left in. The project already has one rule for the first write (first_op_from_status), test-pinned, which treats expired + unique_attempts_remaining >= 1 as retry.

The fallback was changed to use that same rule. A regression test for the Day 21 state (expired, 255 variants left, no next_op → retry) was added to _test_route_first_write.py. Both _test_route_first_write.py and _test_autosolve_fallback.py pass.

With the fix in, one more autosolve.py 21 warmups-day-21 warmups call opened a fresh variant through op=retry and the adapter solved it in 3.3 s.

Lessons

  • A positional ladder is a hard contract. A new day can extend it; the fix is adding the decoder and its position, not guessing from the bytes.
  • One rule for the first write. The project has first_op_from_status; the fallback must call it, so a expired + fresh-variants state picks retry without needing a next_op from the GM.
  • The first ranked Mastery window was lost to the ladder gap; the retry recovered the mode but not the Hall-of-Fame credit. The decoder is now ready for the next time this layer appears.

Files touched