Tactical LogHuntress CTF 2026 // team archive

Days / Day 31 / Warmups / One More Thing

DAY 31//LANE: WARMUPS//STATUS: CORRECT x3

One More Thing

Warmups SSTTMM Modes reported CORRECT in this write-up

Day 31 - One More Thing (warmups) - agent-3

Prompt: "Thirty-two, a little pressure, then sixty-four. Unpack that thought." = base32 -> gzip -> base64 (peel base64, gunzip, base32).

Results

  • Single Solve: CORRECT - FLAG Token2a0ed481ad80fa6ad37a70f04926ef19
  • Time Trial: CORRECT (4.9 s, ranked) - FLAG Token2c0111e5970c53903c4ade871cd5cf83
  • Mastery: CORRECT (4.3 s) - FLAG Token086af78c8d697e0b1fbcd61641f9ddaa

Method

  • Single / Time Trial: one blob, base64 -> gzip -> base32 -> slug; solved by the stock decoder chain with no change.
  • Mastery: 153 numbered-line blobs of 31 rows each (ladder D30 plus a new row 31). Rows 1-30 were already covered. Row 31 = base64 of a gzip stream whose payload is base32 text (e.g. MUWTCOA=) that decodes to the slug tail. The existing _container_texts peeled base64+gzip but stopped at the base32 text, so _structural_unique found no slug and all 153 blobs were skipped (HELD, 0/153 coverage).
  • Fix (clone only): _gzip_b32 decoder, LINE_DECODERS["gzipb32"], LADDER_NAMES_D31 = LADDER_NAMES_D30 + ["gzipb32"], and a (31, LADDER_NAMES_D31) entry first in _KNOWN_LADDERS in scripts/warmups_wip.py. Offline check: 153/153 blobs decode to a valid slug (scripts/test_mm_offline.py), then assignment_runner.py --rerun -> CORRECT.

Lessons + D+1

  • Missing piece: a THREE-layer row (b64 -> gzip -> b32). _container_texts only composes one text codec plus one container layer.
  • Stall: roughly 10 min between the HELD (153 skipped) and the rerun. The first mastery window was a ranked attempt that went gate_blocked; the rerun used a fresh variant (255 left) via op=retry and finished in 4.3 s.
  • Keep in clone vs port: PORT to shared. Make _structural_unique / _container_texts follow up a container result with a second text-codec step (base32/base64/hex), and add LADDER_NAMES_D31 to shared autosolve_warmups.py.
  • D+1: expect row 32 to append one more layer; the title/prompt names the layers in order ("Thirty-two ... pressure ... sixty-four").