One night during the siege of Sung's capital, the Ch'u commander Tzu Fan is asleep in his own camp, behind his own guards, when he wakes to find a stranger sitting on his bed. The man is Hua Yüan, sent out of the starving city. He does not shout for help; there is no help close enough to matter. Hua Yüan says, quietly, that inside the walls the people "exchange their children and eat them, then break the bones for faggots" — and that even so, Sung will never accept a covenant forced on it, even if the state is wiped out. But: pull your army back thirty kilometers, and we will obey.1 Terrified, Tzu Fan swears the oath, reports to his king, and the army withdraws.2 No wall was breached. One man crossed the lines, sat on the right mattress, and ended a siege with a sentence.
That is the stealthy method: get a body into a place no body should be able to reach, then apply pressure — physical or psychological — that the target cannot answer in the moment. The vault files three of these together because Sawyer does, and because they form one technique with rising intensity.
Sawyer groups a "variety of stealthy methods" — the kind used by the "dog thief" in the Lord of Hsin-ling's biography, and the "strong-arm ploys" like the kidnapping of Shih Hui — as a recognizable family of clandestine tradecraft.3 Three cases anchor it.
The first is the coerced covenant. In 680 B.C., at the meeting at K'o, Ts'ao Mei advances on the first hegemon, Duke Huan of Ch'i, sword in hand, and extorts the return of Lu's land on the sunny side of the Wen River. Duke Huan, under the blade, agrees and swears the oath. Afterward Ts'ao Mei throws down the sword and leaves. The striking part is the aftermath: "A covenant so coerced could be contravened, but Duke Huan did not deceive him by doing so." He honored the forced oath — and from that act "Duke Huan's good faith was manifest throughout the realm."4
The second is the night infiltration — Hua Yüan at Tzu Fan's bedside, already described: stealth entry plus the unanswerable threat of a city that will die before it surrenders.5
The third is theft as psychological warfare. The general Tzu Fa of Ch'u, who collected men of unusual skill, took on a self-described thief. When Ch'i attacked and drove Tzu Fa back three times, the thief went to work — by night he cut down the Ch'i commander's bed-curtain and brought it to Tzu Fa, who courteously returned it the next day. The night after, the thief stole the general's pillow; Tzu Fa returned that too. The third night, his hairpin. When Ch'i's troops heard, they were afraid, and the Ch'i commander told his staff: "If we don't leave today, I am afraid the lord of Ch'u will steal my head." He turned the army about and left.6
What unites these is a single mechanism: the demonstration of intimate access. None of the three depends on force in the ordinary sense. They depend on showing the target that you can reach a place he believed was sealed — his dais, his bedside, the pins in his own hair — and on letting his imagination supply the rest.
The escalation across the three is a ladder of how directly the access is weaponized. Ts'ao Mei is the crudest: he is physically present, sword drawn, threat explicit. Hua Yüan is subtler: no weapon, but a body where no body should be and a threat ("we eat our children and will not yield") that makes resistance pointless. The thief is subtlest of all: he never threatens at all. He returns the curtain, the pillow, the hairpin, each politely handed back — and the politeness is the threat, because every returned object is a message: I was beside your sleeping head and chose to take only this. The Ch'i general reads the sequence correctly. The next thing taken could be his head. The thief wins without a single threat spoken, because demonstrated access is a threat the target threatens himself with.7
The coerced-covenant case adds a second axis Sawyer clearly finds important: the question of faith. Duke Huan was under no obligation to keep an oath extracted at sword's point; he could lawfully have repudiated it. He kept it anyway, and the keeping became the foundation of his reputation across the realm.8 So the stealthy method that forces a covenant raises immediately the problem of whether the forced covenant binds — and the tradition remembered Duke Huan precisely because he treated even a coerced word as a real word.
This page hands the vault its clearest case of psychological coercion as tradecraft — leverage applied through access rather than through armies. It is the non-lethal sibling of Assassination Tradecraft Methods: the same skills of penetration and proximity, stopped one step short of the kill. The thief who takes the general's hairpin three nights running is using the exact capability an assassin would use to take his life — and the whole power of the maneuver is that the target knows it.
It also gives Sun-tzu's five-spies framework a category it underweights. Sun-tzu's spies gather information and spread disinformation; these operators do things — extort an oath, lift a siege, panic an army — by physical intrusion. They are the action arm the five-spies taxonomy mostly leaves implicit, and the vault can read them as the operational complement to the intelligence-collection spies.
The thief's three nights deserve close reading because the genius is in the restraint, not the theft. Look at what he takes, in order: the bed-curtain, the pillow, the hairpin.9
The sequence tightens like a noose. The curtain is around the bed — close, but still at arm's length from the sleeping man. The pillow is under his head. The hairpin is in his hair, the most intimate object of the three, touching his body. Each successive theft proves a deeper penetration of the general's personal space, and each is returned, courteously, by Tzu Fa with the same fiction: a soldier gathering firewood happened to find it.10 The fiction is part of the cruelty. Tzu Fa pretends innocence, which forces the Ch'i commander to do the math himself — these were not found in a field; these were taken from my bed, three nights running, by someone who could have cut my throat and instead took a hairpin. The thief never harms anyone. He simply maps, by demonstration, the shrinking distance between himself and the general's life, and lets the general extrapolate the next step. The Ch'i commander's own line completes the proof: he does not fear the loss of the curtain or the pillow; he fears that "the lord of Ch'u will steal my head."11 The operation's payload was never the objects. It was the inference.
You are the thief, and tonight is the third night. You have already returned the curtain and the pillow; you watched, from the dark, as your own general handed each back with that small straight face about firewood, and you understood the game even before he explained it, because he never explained it. Tonight the camp is jumpier — the guards are doubled, the talk in the lines is uneasy — and that is exactly why you go. You move past the doubled guards the way you moved past the single ones, because doubling guards does not help against a man who is already inside their idea of inside. You kneel by the sleeping general. His breath is slow. The hairpin is right there, in the knot of his hair, and your fingers close on it as gently as on the pillow last night, as gently as you would on his throat. You take only the pin. You leave the throat. You leave it on purpose, and you make sure, by leaving it, that he will spend tomorrow knowing you left it. Then you are gone, and the pin is in your hand, and somewhere across the camp a man who outnumbers your general is about to decide to go home.
The method rots when the demonstration of access is mistaken for the goal. The thief who actually kills the general has thrown away the weapon — a corpse cannot panic an army the way a returned hairpin can. The power lives entirely in the withheld harm; spend the harm and you forfeit the leverage. The diagnostic sign of failure is the operator who escalates to violence because intimidation feels incomplete. It was already complete.
The second failure is the unhonored covenant. Sawyer's whole reason for telling the Ts'ao Mei story is the road not taken: Duke Huan could have broken the coerced oath and lost nothing legally, but he would have lost the realm-wide reputation that made him hegemon.12 The coercer who extorts an oath and then watches the target repudiate it has gained a worthless scrap of paper and taught the world he can be defied. The stealthy covenant only "works" if the target's own commitment to faith — not the coercer's force — keeps it. That is a control the operator does not hold.
The third sign is the security that learns. Sawyer notes the night-infiltration cases say "as much about lax security measures... as the thief's skills."13 These operations exploited camps that should have been locked down by the military writings' own standards. A target who actually tightens — who makes the bedside genuinely unreachable — collapses the entire technique, because demonstrated access is impossible where access is impossible.
Sawyer is openly skeptical of the Ts'ao Mei episode's plausibility. He notes that Ts'ao Mei's lone determination, "although perhaps formidable, could have been quashed by Kuan Chung or armed attendants who must have been near the dais," and that the story "struck the imagination of later listeners because of its improbable nature."14 So the founding case of the coerced covenant may be more legend than event — preserved because it was improbable and morally striking, not because it was typical. The vault should hold it as exemplary lore, "much referred to in subsequent military writings," rather than as reliable operational history.15
A related tension: Sawyer flags that Ts'ao Mei "never directly threatened Duke Huan with harm," and treats that as "an ameliorating aspect that granted the incident a degree of respectability" — distinguishing it from the cruder threats in the cases that follow.16 This means the family of "stealthy methods" is not morally uniform in the tradition's own eyes. The forced-but-unthreatening covenant was respectable; the bedside extortion under threat of cannibal desperation was not. Sawyer preserves that gradient rather than collapsing the cases into one tactic.
The open question is the security one, which Sawyer raises and leaves: do these stories teach us about brilliant infiltration, or about how badly even well-warned commanders kept their own camps?17 The military writings preached lock-down; the famous cases all turn on lock-down failing. The vault cannot resolve whether the technique was genuinely hard to stop or merely exploited a discipline nobody actually maintained.
Sawyer the storyteller and Sawyer the analyst pull against each other across this passage. As storyteller he relishes the drama — the bed, the sword, the three nights — and these are among the most vivid scenes in his book. As analyst he keeps puncturing them: Ts'ao Mei should have been cut down; the infiltrations prove lax security as much as skill; the K'o covenant is "improbable."18 The convergence he lands on is subtle: these episodes survived because they were improbable and morally legible, which means the historical record of "stealthy methods" is curated by memorability, not frequency. The tradition kept the stories that made a clean point — faith honored, fear weaponized — and we cannot tell from them how routine or how rare such operations actually were. Sawyer's continuity claim, that these illustrate a coherent body of "stealthy methods" running through Chinese practice, is therefore an inference from a curated sample, and the page marks it as such.
To Assassination Tradecraft Methods. These are the same skill set divided by a single decision — to harm or to withhold — and reading them together exposes what assassination cannot buy. The infiltrator and the assassin share everything technical: the penetration of a guarded space, the approach to a sleeping or seated target, the proximity to the body. The assassin spends that proximity on a death. The stealth-coercer banks it, and the banked proximity turns out to be a more flexible instrument than the spent one — the thief who takes the hairpin gets an entire enemy army to retreat, an outcome a single assassination could not reliably produce. The structural insight: in the Chinese covert repertoire, demonstrated lethality is often worth more unspent than spent, because a living, terrified commander makes the decision you want, whereas a dead one is replaced by someone you have not yet reached. Neither page alone tells you that restraint can be the higher-yield use of an assassin's exact capability — that the most efficient assassination is sometimes the one you visibly decline to commit.
To the Lord of Hsin-ling's Tiger-Tally Theft. Sawyer explicitly links these stealthy methods to the "dog thief" of the Lord of Hsin-ling's biography, and the two pages together reveal a structural unity in how the tradition valued the man of low skill.19 The tiger-tally theft turned on a thief and an impersonator infiltrating to steal the very token of command authority; the bedside and hairpin operations turn on infiltrators reaching the body of command itself. In both, the decisive operator is a social outcast — a thief, a "brigand of the realm" whom Tzu Fa's own attendants despise — whose disreputable skill becomes a strategic asset no general possesses.20 The tension the handshake surfaces is between the military writings' insistence on order, hierarchy, and lock-down, and the repeated historical fact that wars were turned by men outside that order, doing precisely what the order was built to prevent. The insight neither page generates alone: the Chinese strategic tradition simultaneously preached perfect camp security and cherished the stories of the thieves who defeated it, because the thief proves that the gap between doctrine and reality is exactly the space where covert operations live. The tiger-tally page shows the token of authority stolen; this page shows the body of authority reached. Both say the locked door was never really locked.
Sharpest implication: The most powerful thing an infiltrator can do is leave — to reach the target's throat and visibly take only a hairpin — because withheld lethality coerces a living decision-maker more reliably than a death coerces his replacement.
Generative questions: