Tactical LogHuntress CTF 2026 // team archive

Days / Day 30 / Warmups / Percent Percent

DAY 30//LANE: WARMUPS//STATUS: CORRECT x3

Percent Percent

Warmups SSTTMM Modes reported CORRECT in this write-up

Percent Percent (warmups, Day 30)

Layer of the day: percent/URL encoding applied TWICE (%25%37%32 -> %72 -> r).

Results

  • single_solve: CORRECT (2.0 s) - FLAG Tokenc3470492441834ccc8a5034caef50c61
  • time_trial: CORRECT (5.9 s) - FLAG Token279b2bc483500c0528caef2c29fffcd2
  • mastery: CORRECT (6.1 s, 148 sets) - FLAG Token541b6a7b26e311585c0e97eff07ec8a6

Single / Time Trial were solved by the existing peel (double percent handled). Example slug from the first Mastery attempt's set1: dapper-chimney-juniper-30.

Mastery structure (30 numbered lines per blob)

Lines 1-23 follow the known ladder (base64, rot13, hex, binary, decimal, octal, percent, html, json-u, base32, base64, reverse, atbash, caesar7, rot47, morse-hex, bacon-hex, rail3, utf16le-hex, utf16be-hex, gzip-b64, zlib-b64, ascii85). Slug region: 24 RLE N:C,, 25 U+hhhh codepoints, 26 brainfuck, 27 Ook, 28 reversed hex, 29 rot13+base64, 30 NEW: percent-encoding applied twice.

Fix (solver clone only: scripts/warmups_wip.py)

  • Added LINE_DECODERS["percent2"] (unquote twice) and LADDER_NAMES_D30 = D29 + ["percent2"].
  • _guided_positional selects the D30 ladder when a blob has exactly 30 lines.
  • Offline check against the 148 downloaded blobs: 148/148 decoded, all slugs match the three-words-NN shape.

Lessons + D+1

  • Missing piece: a 30-line blob fell back to the 21-name ladder, so the slug-region lines 28-30 had no trusted decoder and the whole blob returned None. First Mastery start ended HELD with 148 skipped blobs (the ranked attempt was gate_blocked, then the untimed fallback was spent); a rerun after the fix used op=retry and a fresh variant (255 left).
  • Stall: about 10 minutes from HELD to the fixed rerun, mostly reading the guided decoder.
  • Keep in clone vs port: PORT to the shared autosolve_warmups: generalize ladder selection (extend the previous ladder by one name per extra line, instead of a len== check per day) and register percent2.
  • D+1: expect the ladder to grow by one line per day again; the next line is likely another double/nested encoding on the slug tail. Run _testall.py-style offline check against the Single's structure before the timed start.