Indra is about to fight the asuras and is given a piece of equipment. It is not a weapon and not a shield. It is a procedure: a sequence of syllables installed on specific parts of the body, followed by a long prayer that names a different form of Viṣṇu for every place a person might be, every kind of harm a person might suffer, and every division of the day and night.
Read as devotion, it is a hymn. Read as engineering, it is a threat model — an attempt to enumerate the space of dangers and assign a defender to each cell.
That reading is what makes it worth a page, because enumeration is exactly the design this volume elsewhere shows failing. And because the chapter ends with a worked example in which the armour harms somebody it was never aimed at.
Before any protection is claimed, there is a setup procedure, and it is specified to a level the rest of the chapter does not attempt.
The practitioner washes hands and feet, performs the ablutions, wears a ring of kuśa grass, sits facing north, controls his speech, and purifies his limbs with two mantras. Then the eight-syllable formula is apportioned "progressively" across the two feet, two knees, two thighs, stomach, heart, chest, mouth and head — and then "Oum namo Narayanaya is then repeated in the reverse order."1
Then a hand-installation using the twelve-syllable mantra: "It starts with the index finger and ends with the joints of the thumbs." Then a second placement — Oṃ in the heart, vi at the crown, śa between the eyebrows, na in the tuft of hair, ve in the two eyes, na in all the joints, and ma held "as a weapon."1
The stated outcome of the installation is not protection but transformation: "a learned person thus becomes a personified form of the mantra." The practitioner is not wearing the armour. He has been rebuilt as it.
The prayer proper covers the threat space three times over, along three different axes.
By place. Matsya in the waters, against aquatic creatures and Varuṇa's noose. Vāmana on the land. Trivikrama in the sky. Nṛsiṃha "in impenetrable places, forests and the forefront of the battle." The sacrificial form along roads. Paraśurāma on mountain summits. Rāma with Lakṣmaṇa in foreign lands.1
By kind of harm. Fifteen assignments, and the matching is not arbitrary — each avatāra is given the danger his own story qualifies him for:1
| Defender | Threat |
|---|---|
| Nara | pride |
| Dattātreya | deviations in yoga |
| Kapila | the bondage of karma |
| Sanatkumāra | the god of love |
| Hayaśīrṣa | ignoring gods along the road |
| Nārada | transgressions in worship |
| Kūrma | all kinds of hell |
| Dhanvantari | eating undesirable food |
| Ṛṣabha | the opposite sentiments |
| Yajña | hardship caused by other people |
| Balarāma | the krodhavaśas |
| Vyāsa | ignorance |
| Buddha | heretics |
| Kalki | the impurities of the kali yuga |
| Nārāyaṇa | adharma and distraction |
The celibate handles desire, the physician handles bad food, the compiler of the Vedas handles ignorance. The avatāra's biography is the lookup key.
By time. Twelve watches, each with its own name and, in the daylight hours, its own weapon: Keśava with the club in the morning, Govinda with the flute at saṅgava, Nārāyaṇa with the spear in the forenoon, Viṣṇu at midday, Madhusūdana with the bow in the afternoon, Mādhava in the evening, then Hṛṣīkeśa, Padmanābha, the lord of the śrīvatsa, Janārdana, Dāmodara, and finally "the illustrious Vishveshvara, whose form is time."1
The cycle closes on the deity of time itself, which is the only entry that names the axis it sits on.
There is a fourth layer, addressed in the vocative to the weapons directly — cakra, club, conch, sword, shield. What is notable there is the qualifier: the cakra is asked to burn "all the enemy soldiers who deserve to be burnt down," the club to crush "those who should be crushed down."1 The instruments are given discretion criteria, as though the danger of an indiscriminate defence had occurred to somebody.
The chapter then supplies a worked example, and it is not a story about winning a battle.
In ancient times, a brahmana named Koushika mastered this. Using the path of yoga, he cast off his life in the desert regions. Chitraratha, the lord of the gandharvas, surrounded by women, once travelled in a vimana, over the spot where the brahmana had died. He instantly fell down from the vimana in the sky, with his head facing downwards.1
Attend to who this happens to. Citraratha is not an asura, not an enemy, not a threat to anyone. He is a gandharva king out flying with his retinue, and he does not know Kauśika is there — Kauśika is dead. He is knocked out of the sky head-first for passing over a grave.
Three properties of the armour follow from this, none of them stated by the chapter:
It outlives the wearer. Kauśika is dead and the effect is undiminished.
It is attached to the physical residue. The remedy is not a prayer or an apology. Citraratha, advised by the Vālakhilyas, "collected the bones and immersed them in the Sarasvati, which flows in an eastern direction."1 The bones are the hazard, and disposing of them clears the site.
It cannot tell an enemy from a passer-by. This is the decisive one. The armour was specified against aquatic creatures, asuras, heretics, bad food and the kali yuga. It has no representation of Citraratha at all — and in the absence of one, it treats overflight as an attack.
So the chapter's one demonstration of the kavaca in operation is a false positive with a third-party casualty, presented without comment as evidence of the armour's power. It is offered as reassurance. It reads as an incident report.
The volume runs the design a second time, in a completely different register, at 10(6).
After Pūtanā's body has been burnt, the gopīs perform nyāsa over the infant Kṛṣṇa. The structure is immediately recognisable: a body-part partition — Aja at the feet, Maṇimān at the knees, Yajña at the thighs, Acyuta at the loins, Hayāsya at the stomach, Keśava at the heart, Īśa at the chest, Ina at the throat, Viṣṇu at the arms, Urukrama at the mouth, Īśvara at the head. Then a directional partition — the cakra-wielder in front, Hari with the club behind, bow and arrow on both sides, the conch-bearer at the corners, Upendra above, Tārkṣya on the ground. Then a demon list of eighteen or more classes.2
But the third axis has been swapped. Where 6(8) partitions the day into twelve watches, 10(6) partitions activity: Govinda in his pastimes, Mādhava while sleeping, Vaikuṇṭha while walking, the lord of Śrī while seated, the enjoyer of sacrifices while eating.2
The substitution fits the subject exactly. An adult general travels through domains and fights at hours; an infant does not go anywhere and has no schedule, but he does sleep, walk, sit and eat. The architecture is identical and the partition is fitted to the protected party's mode of life. That is a real piece of design, and it is the strongest evidence that the kavaca form is a genre with rules rather than a one-off hymn.
Two things about the scene are left unremarked by the text.
The rite is performed after the attack. Pūtanā is already dead; the gopīs stood "stupefied" while she died. And the list they recite includes pūtanās as a class of demon.2 The category is invoked once the individual has been dealt with.
And the being it is installed on is the source of every name in it. The gopīs are protecting Kṛṣṇa by invoking Viṣṇu, Acyuta, Keśava, Govinda, Mādhava and Vaikuṇṭha — over Kṛṣṇa. He had just killed the demon himself, unaided, with his eyes closed.
The kavaca is never defeated in this volume. Indra wears it, wins, and "enjoyed the prosperity of the three worlds."1 Nothing here should be read as the text staging its failure.
What can be said is structural. This build has found six instruments in this volume that protect their holder and fail at unenumerated cases — boons whose terms are exhaustive lists, defeated at the seam between the listed items. The kavaca is built on precisely that principle: seven places, fifteen harms, twelve watches, eighteen demon-classes. Its coverage is its enumeration.
The chapter's own closing claim reaches past what its lists can support — a wearer "does not suffer any fear from any direction — from kings, bandits, evil planets, or physical or mental diseases."1 Any direction is a universal quantifier over a defence assembled from particulars, and the same sentence immediately narrows it to four examples.
The volume does not test this, and the honest statement is that the kavaca shares its architecture with the instruments that fail rather than that it fails. The one anomaly the chapter does record is not under-coverage but over-firing — Citraratha, who was in no listed category and was struck anyway.
Separate the installation from the claim. The setup at 6(8) is specified to the finger-joint; the coverage claim is any direction. Where a procedure's precision and its promise are at different resolutions, the promise is not derived from the procedure.
Read a protective enumeration for its partition, not its length. Place, harm-type and time in one version; body-part, direction and activity in the other. What a threat model divides the world by tells you who it was built for, and the swap between the two versions is the clearest evidence of intent in either.
Look for the worked example and check who it happens to. The kavaca's demonstration involves no adversary. A defence illustrated by a bystander casualty is telling you something about its discrimination that its specification does not.
Note when a protective rite is performed after the event. 10(6) is retroactive, and the class of demon it names had already attacked. That is a fact about the rite's function — consolation, ratification, or routine — which the enumeration itself cannot show.
Coverage is achieved by listing. Seven places, fifteen harms, twelve watches. A defence whose completeness is an inventory is exposed at every case not on it.
The defenders are matched by biography. Sanatkumāra against desire, Dhanvantari against bad food. Where assignments track the assignee's story, the model is being built by association rather than by analysis of the threat.
A universal claim sits beside a short list of examples. No fear from any direction — from kings, bandits, evil planets, or diseases. The narrowing in the same sentence is the tell.
The instruments are given discretion criteria. Those who deserve to be burnt down. Someone has anticipated over-firing and addressed it with a qualifier rather than a mechanism.
The effect persists in a physical residue. Bones that must be disposed of. A protection that requires decommissioning is an object in the world, not a state of the protected person.
Strongest support. The installation sequence, all three partitions, the fifteen threat-assignments, the twelve watches, the weapon-vocatives and the Kauśika episode are quoted or tabulated from 6(8); the body-part, direction and activity partitions and the demon list from 10(6). The two are compared on their own terms without importing any external account of nyāsa.
Tension the page preserves. Whether Citraratha's fall should be read as the armour malfunctioning or as it working correctly — the site being genuinely sacred and overflight genuinely an offence — is not settled by the chapter. The text presents it as a demonstration of power; this page reports that the person harmed is in none of the armour's categories, and does not adjudicate.
Absence-claim discipline, and it matters here. The kavaca is never defeated in this volume is a claim about the passages read in this build, principally 6(8), 10(6) and the battle narratives of Skandhas 6 and 8. It is not a claim about the corpus, and the vault should not treat it as one. The standing rule applies: an absence asserted over unread material is invention.
On the seam-defeat connection, stated at the right strength. The kavaca is architecturally an enumerative defence and this build's seam-defeat family are demonstrably defeated enumerations. That is a shared design, not a shared outcome. No passage shows the kavaca failing at a seam, and this page does not claim one does.
Open question. Do other Purāṇic kavaca texts partition on the same three axes, and does the partition track the protected party's mode of life there too? One further instance with a fitted partition would move that observation from a described pair to a rule of the form.
The volume holds two incompatible accounts of how a person is kept safe and does not reconcile them.
One is this: an installed, ordered, exhaustively specified apparatus, requiring a north-facing seat and correctly placed syllables, whose effect is strong enough to persist in a corpse and knock a stranger out of the sky. The other is the account this corpus is famous for — Ajāmila calling his son's name, meaning nothing by it, and the entire machinery of punishment standing down.
Both are in this volume. Neither is presented as the lesser. And the difference is not one of degree: one demands total specification and the other demands nothing at all, not even intention.
There is a further convergence with the decommissioning material at 7(12). Both chapters treat a person's constituents as a set of positions with divine occupants — there, fifteen faculties returned to their issuers; here, eleven body-parts and twelve watches assigned their defenders. Installation and de-installation use the same map. The kavaca puts a god on each part; the protocol hands each part back to a god. That the volume runs the map in both directions, in two Skandhas, and never once notes the symmetry is by now the most reliable thing about it.
JACA Threat Assessment Framework — model the actor or model the world, and inherit opposite failure modes.
De Becker's JACA assesses a person: does he feel justified, does he perceive alternatives, does he fear consequences, does he have the ability. Four questions about one adversary's decision process. The framework carries no list of dangerous situations at all.
The kavaca does the reverse and carries no adversary model whatever. It enumerates the world — seven places, fifteen harms, twelve watches, eighteen demon-classes — and never asks who might want to cause harm or why.
The insight neither produces alone: the two designs fail in mirror-image ways, and the failure mode is a consequence of what got modelled. An actor model generalises to adversaries never met, because the four questions can be asked about anyone; it fails when the actor is unreadable — no history, no linguistic sample, no access. A context model needs no access to anyone; it fails at any situation not on the list, and — this is the part JACA's framing makes visible — it cannot distinguish an attacker from a bystander, because distinguishing them requires exactly the actor-assessment it omitted. Citraratha is that failure in narrative form: a defence with no representation of intent, treating a man flying past as a man attacking.
Neither framework can see its own gap from the inside, which is the practical yield: a threat model built on contexts should be audited by asking who it thinks the adversary is, and one built on actors by asking where it assumes they will be found.
Sharpest implication. The one time this volume shows the Nārāyaṇa-kavaca operating outside a battle, it kills nobody's enemy — it knocks a gandharva king out of the sky for flying over a dead practitioner's bones, and the fix is to dispose of the remains. A defence with no model of intent cannot tell an attack from a transit.
Generative questions.