One possibility for the language or language family that once included Harappan is that it continues to survive today in the form of a language isolate. (Previous analysis at this blog here).
If that were the case, the Burushaski language would be the most plausible candidate, by virtue of location (in Kashmir), established antiquity (from references in old Tibetan texts), due to the presence of retroflex consonants, and because all other plausible candidates (Indo-European, Dravidian, Munda, and Uralic) can be ruled out. Pre-fixing was probably part of the Harappan language and is present in Burushaski as well, although it also has suffixes and doesn't necessary pre-fix in the same way. The fact that Burushaski has some ergative features and that it makes an animate/inanimate distinction also supports the hypothesis, although it does not apparently use the k- prefix to make that distinction.
There is one other language isolate in India, the Nahali language, but its location in West Central India is less of a good fit, and its antiquity is in doubt (Nahali may have a "thieves' cant" with roots in a Munda language developed in the 1800s for the purpose of being indecipherable to outsiders.) It also has retroflex consonants (an areal feature of many South Asian languages, likely derived from contact with Harappan).
As Wikipedia notes:
The Harappan language is the unknown language or languages of the Bronze Age (2nd millennium BC) Harappan civilization (Indus Valley Civilization, or IVC). The language being unattested in any readable contemporary source, hypotheses regarding its nature are reduced to purported loanwords and substratum influence, notably the substratum in Vedic Sanskrit and a few terms recorded in Sumerian cuneiform (such as Meluhha) . . . There are a handful of possible loanwords from the language of the Indus Valley Civilization. Sumerian Meluhha may be derived from a native term for the Indus Valley Civilization, also reflected in Sanskrit mleccha meaning foreigner and Witzel (2000) further suggests that Sumerian GIŠšimmar (a type of tree) may be cognate to Rigvedic śimbala and śalmali (also names of trees).
I don't have any stronger evidence in favor of the hypothesis, but because other possibilities seem unlikely, it deserves more attention. Presumably, efforts have been made to show lexical connections that have failed to provide strong evidence of a connection, but I don't know that for a fact.