Sanskrit dramatic theory has a word for an envoy who is trusted to work out what the mission requires.
Niśṛṣṭārtha dūta — an envoy sent to act on his own judgment rather than to relay a fixed message, free to persuade, warn, entice or divide as the moment demands, in service of a goal his master never spelled out in detail.1
It is a strange and useful category, and its strangeness is worth sitting with. The defining feature is not competence or loyalty. It is that the brief is incomplete on purpose — the sender has a goal, has not articulated it fully, and accepts that what the envoy does in the room will be determined by the room.
Every organisation has this role and almost none names it. That is what makes the term worth having.
Devdutt Miśra's evidence is not the famous mission to Laṅkā. It is a much earlier scene, and choosing it is the essay's best move.
At Ṛṣyamūka, Sugrīva sends Hanuman to do one thing: find out who these two armed strangers are. That is the whole brief.
What Hanuman does instead is notice a structural parallel nobody has pointed out to him — Sugrīva has lost a kingdom and a wife to Vālin; this stranger has lost a wife and is looking for allies — and propose an alliance that nobody asked him to broker.1
The mission he was given was reconnaissance. The mission he performed was diplomacy, and it produced the alliance the entire remainder of the epic runs on.
That is the category in its pure form. He was not exceeding his authority in a way that needed forgiving; he was exercising the discretion the role assumes. Sugrīva's actual goal — survive Vālin — was never stated in the brief and was correctly served by ignoring it.
The volume prints the contrary position immediately before, and does not notice.
Sītārām Caturvedī's envoy essay ends at p. 193; Miśra's begins there. Caturvedī's requirement is that an envoy be scrupulously accurate in reporting back only what was actually said, and he praises Hanuman for carrying messages word for word, adding nothing and softening nothing.2
Set the two beside each other and they are not complementary virtues.
Caturvedī's envoy is a conduit. Success is measured by fidelity, and every departure is loss. Miśra's envoy is an agent. Success is measured by whether the master's unstated goal advanced, and strict fidelity to an inadequate brief is failure.
The same act scores oppositely. Hanuman at Ṛṣyamūka, brokering an alliance he was not sent to broker, is Miśra's exemplar and is a straightforward breach of Caturvedī's standard.
Miśra defends two of Hanuman's choices at Ṛṣyamūka, and the defences show what the category licenses.
The Brahmin disguise. Miśra treats it as sound tradecraft — it lowers the threat the strangers perceive and lets him observe their unguarded reactions. He supports it with the Kirātārjunīya, where Yudhiṣṭhira sends an agent disguised as a renunciant to assess how Duryodhana treats his subjects.1
The bow, which creates a protocol problem: a Brahmin does not bow to kṣatriya householders. Miśra's defence is the Mahābhārata scene where the assembled nobles rise for Kṛṣṇa in Duryodhana's court, in defiance of an explicit prohibition, because his presence commanded it.1
Notice what the two defences have in common. Both concede that a rule was broken and argue the breach was correct. The disguise is deception; the bow violates the very identity the disguise established.
That is the category doing real work. A conduit-envoy has no authority to deceive or to break protocol, because neither is in the brief. A niśṛṣṭārtha dūta does — and the price is that his conduct can only be judged after the fact, by outcome, which is precisely what makes the role dangerous to grant and useless to specify in advance.
Discretion granted without a check is just an unsupervised agent, and the volume supplies the check — in a different essay, by a different contributor, who does not appear to know he is answering Miśra.
Lakṣmaṇ Śāstrī's evidence for abhyastakarmā, seasoned judgment, is the Trikūṭa pause. Hanuman reaches Laṅkā before sunset and does not enter. He climbs the peak, conceals himself, and watches — reasoning that envoys who trust their own cleverness ruin missions, and that acting against the right place, time and circumstance undoes even well-planned efforts. He begins searching only after nightfall.4
That is the same figure, in the same mission, exercising discretion by declining to act.
And the reasoning is a direct constraint on the category. Niśṛṣṭārtha dūta licenses initiative; the Trikūṭa principle says the licence is most dangerous in the hands of somebody enjoying it. The check is not a rule about what an envoy may do — it is a disposition about when to move, and it can only be held by the same person holding the discretion.
Which is the honest shape of the thing. Discretion cannot be constrained from outside without becoming a brief again, so the tradition's answer is to make the constraint part of the qualification: the envoy you can send unbriefed is the one who has already learned that his own cleverness is the failure mode.
The volume never asks the question, but its own material answers it, and the answer is not about the envoy.
It is about whether the sender knows what they want.
Sugrīva at Ṛṣyamūka does not. He wants to survive Vālin and has no plan for it; the brief he issues is the brief of a man who cannot specify the outcome he needs. Discretion is the only thing that can serve him.
Rāma sending the signet ring to Sītā knows exactly what he wants: that she receive this object and this message and believe them. There is nothing for judgment to add, and every deviation is damage.
So the two doctrines are not rivals. They are answers to different questions, and the question is how well-specified the goal is at the moment of sending. A well-specified goal wants a conduit. An under-specified one wants an agent — and the sender's inability to say what they want is the condition that creates the need, which is why the choice cannot be made by assessing the envoy.
This is not an antiquarian point, and the extension corpus this build also reads stumbles into it without the vocabulary.
Its AI-collaboration material proposes "absolute fidelity to the original transmission while adapting the communication for optimal reception" as the model for working with a system.3 Both halves at once — which, as the volume demonstrates across two consecutive essays, is a tension rather than a synthesis.
The tradition names both ends and the corpus names neither. Niśṛṣṭārtha dūta is the discretionary pole; Caturvedī's word-for-word is the fidelity pole; and a practitioner deciding what to delegate is choosing between them rather than blending them.
Which yields the operative question, and it is not how capable is the delegate: can I specify what I want? If yes, delegate as a conduit and measure fidelity. If no — and the honest answer is often no — then discretion is what is actually needed, and the thing to build is not a tighter brief but a way of judging outcomes, since that is the only judgment available.
Before delegating, ask whether you can state the goal completely. Not the task — the goal. If you can, you need fidelity, and the brief should be exact and the review should check for drift. If you cannot, you need discretion, and a brief pretending to completeness will produce the worst outcome available: an agent constrained to a specification you already know is inadequate.
Name which kind of delegation this is, out loud, when you hand it over. Most delegation failures are not capability failures; they are two people holding different theories of the role. The sender expects a conduit and receives initiative, or expects initiative and receives literal compliance — and both parties performed correctly by their own lights.
And accept the cost of discretion honestly: you cannot specify it, so you can only judge it afterwards. The tradition's own defences of Hanuman's disguise and bow are both retrospective. There was no rule that permitted them in advance and no rule could have been written. If you are not willing to assess by outcome, you are not actually granting discretion — you are granting it and reserving the right to punish it, which is worse than withholding it.
Two Aṅka essays in the Mahavir Aaradhyaa edition,12 and one document of the extension corpus, cited only as evidence about itself.3
A sourcing weakness, stated plainly. Miśra attributes niśṛṣṭārtha dūta to Sanskrit dramatic theory's typology of messengers and does not name the treatise.1 The category is genuinely attested in that literature, but this build has not verified the definition against a printed source, and the essay supplies no locator. [TRADITION-CLAIM — unverified] on the definition's exact wording.
Bounded. The claim that Miśra names no other envoy category is a claim about this essay. The volume's four envoy essays between them name no shared typology at all, which is treated separately.
Open questions. Which text carries the typology — the Nāṭyaśāstra, the Daśarūpaka, or a nīti source? And what are the other categories? A typology with one named member is not a typology, and the neighbouring terms would sharpen the boundary considerably — a category defined only by its most permissive member cannot show where discretion stops.
Miśra's method is worth naming because it is unusual in this volume and cuts both ways.
He reconstructs Hanuman's reasoning — what he must have noticed at Ṛṣyamūka, why he chose the disguise, why he bowed — and the essay is explicit that this is inference rather than a text's account. Nothing in the Rāmāyaṇa narrates the deliberation.
That makes it weak as evidence and strong as analysis. He is not reporting what Hanuman thought; he is demonstrating that the behaviour is intelligible as strategy, which is a different and more defensible claim, and one the reader can assess directly.
Against Caturvedī the contrast is instructive: Caturvedī lists qualities and matches episodes to them, while Miśra takes one episode and asks what kind of role it implies. The second method produces a category that transfers. The first produces a portrait.
Fidelity and Adaptation: The Messenger Model. That page has the extension corpus's formulation — fidelity plus adaptation — and the Aṅka's supply of the missing variable, the vernacular choice calibrated to what a frightened listener can believe. This page supplies the other half: the tradition does not merely resolve the tension case by case, it has a name for each pole. What the pairing produces is that the corpus's formulation is not a synthesis but an unrecognised choice presented as a single instruction — and an instruction to do two incompatible things reliably produces whichever the practitioner was going to do anyway. The repair is not better balance; it is asking, per task, which pole applies.
The Catalogue of Singular Trust. That page sorts the seven assignments given to nobody else and finds four of them are messages, none is a battle, and the category is tasks that fail completely if the carrier garbles them. Set against this page, the two produce a delegation grid rather than a single test. One axis is degradation — does partial failure yield a partial result. The other is specification — can the goal be stated. The four combinations are genuinely different jobs, and the epic uses Hanuman for two of them: the ring to Sītā is high-degradation and fully specified, needing a conduit whose fidelity is verifiable; Ṛṣyamūka is low-degradation and unspecified, needing an agent. Sending the wrong kind into either is the ordinary way delegation fails.
The uncomfortable part is that discretion is only visible as a category when it succeeds.
Hanuman's improvisation at Ṛṣyamūka is celebrated because the alliance worked. Had it failed — had Rāma taken offence at a disguised stranger, had Sugrīva been compromised by the approach — the same conduct would be recorded as an envoy exceeding his brief, and Miśra's category would not have been reached for.
Which means the tradition's examples are selected on the outcome, and a reader who takes niśṛṣṭārtha dūta as a licence is taking it from a sample containing no failures. The category is real; the evidence for granting it in any particular case is not in these essays, and could not be.