Ashes to Assets (Day 25, script-analysis)
GM id: script-analysis-day-25-phoenix-stage-chain
Flags (GM-accepted)
| Mode | Result | Flag |
|---|---|---|
| Single Solve | CORRECT (autosolve at claim, 1571 fields) | Tokene241479c34c84073129d3c786d934808 |
| Time Trial | CORRECT (1625 fields, regular path) | Token9a93079f29da4138f4cbb4ecd49478f7 |
| Mastery | CORRECT (ranked, 14625 fields) | Tokena8f19a3101355a4450b53edc77896a61 |
Challenge
AES zip evidence.zip (challenge artifact, not published) (password infected), 167 inert fixtures in ~26 languages (ps1, js, vbs, bat, py, sh, php polyglots, Elixir/Phoenix host, JSP, Lua, Ruby, ...). Every fixture holds a Material_<id> / Codec_<id> pair that decodes into a v=N|key=value|... descriptor. Questions ask for role, ATT&CK id, steps, campaign, indicator, carrier, obfuscation, and so on.
What blocked the Time Trial and the fix
- v=2 descriptors. The Time Trial variant added hash-named Python fixtures (
3db3db.py, ...) whose descriptor beginsv=2|.decode_bash.decode_pyalready decoded them (codecshex_escape|single_byte_xor,urlsafe_base64|repeating_key_xor,base64_chunks|single_byte_xor,integer_tuple|rolling_xor). Butautosolve_script.generic_decodeand_raw_decodeaccepted onlyv=1|, so 14 ids (indicator / steps / campaign-token) were left unresolved. The fix accepts^v=\d+\|everywhere inchallenges/_lib/autosolve_script.py. - Pipe-pair fields.
python-<hash>-carrierand-obfuscationask for "two labels joined by | in that order" (carrier|windows_chain, syntax|obfuscation). An earlier builder set only the single label, and latersetdefaultpasses never replaced it. The first full submit was graded incorrect (all-or-nothing). Fix: a post-pass insolve_fields_extre-derives every id whose question says "two labels joined by |" throughsolve_poly_d8._d23_fill, which joins the pair. This affected 26 fields.
An audit of every answer against its own question (scripts/audit_tt.py, scripts/templates_tt.py) came back clean. Autosolve then got Time Trial CORRECT and Mastery CORRECT in a single rerun.
Example decode (3db3db.py, hex_escape|single_byte_xor|31): v=2|steps=profile_host,acquire_remote,decode_embedded,write_file,dispatch_process|adapter=cpython_loader_handoff|...|campaign_id=campaign-b4c8cc7553|url=https://b4c8cc7553.example.invalid/a14d82c1.py|... gives indicator = URL, steps = profile_host>acquire_remote>decode_embedded>write_file>dispatch_process, token b4c8cc7553, carrier embeddable_python_runtime|direct_python_exe.
Process notes
- The Time Trial was first HELD because a stale attempt existed and the client refused
op=retry. It was run withHUNTRESS_TIMED_START=commander-overrideafter the user explicitly authorized continuing. One incorrect Time Trial submit happened before fix 2. - Lesson: on timed routes, check every answer against the question's own wording ("joined by |", "three labels") before submitting. Single-label answers from older builders can silently win.
Scripts: scripts/diff_v2_patch.py, scripts/audit_tt.py, scripts/templates_tt.py.
<!-- kb-check not-a-script: 3db3db.py, a14d82c1.py -->