# Antagonists — six recurring nemeses **Built 2026-08-24.** The middle weight of the hostile stack ([`../../docs/npc-containers.md`](../../docs/npc-containers.md) §8): heavier than a dropped threat, lighter than a front. > ## The one thing this directory is for > > ### *"Oh. It's him again."* > > That reaction requires the player to hold **all of them at once**. Six is not > a starting number — it is the ceiling. Twenty pre-built nemeses is twenty > strangers. --- ## The six | # | axis — what they want | who | chassis | exit | attaches to | |---|---|---|---|---|---| | 1 | something **from** the party | [Maelis Corr](maelis-corr.md) | Wizard 3 *(Abjurer)* | resolved | Zephyr / Llewellyn | | 2 | the party **stopped** | [Ilma Danneke](ilma-danneke.md) | Bard 3 *(Lore)* | turned | **party-wide** | | 3 | the party **destroyed** | [Corwen Hale](corwen-hale.md) | Fighter 3 *(Battle Master)* | **killed** | Jornis | | 4 | the party **recruited** | [Ossa Rell](ossa-rell.md) | Wereboar *(SRD, CR 4)* | turned | Zephyr | | 5 | **does not know** they are one — and is **wrong** | [Teodor Wick](teodor-wick.md) | Scout *(SRD)* — not a class build | resolved | **party-wide** | | 6 | **is right** | [Iselde Varr](iselde-varr.md) | Rogue 3 *(Thief)* | turned | Deerx | **Secrets files, where earned:** Maelis, Ossa, Iselde. The other three carry their hidden material in `Knows` with `visibility: true`, per the container spec — a separate file is earned, not automatic. ### Why the axis, and not "social / martial / arcane" Class variety changes **what dice get rolled**. For someone the player meets five times, what has to vary is **what scenes they generate**. Six vendettas would be one nemesis with six faces — and that is the failure mode a small bench is most exposed to. **5 and 6 are a deliberate pair.** Teodor is sincerely wrong; Iselde is entirely right. Running both is the whole spread, and neither works nearly as well alone. --- ## The three rules ### 1. They accumulate and never reset By the fourth meeting they know the party fights at range and they are **still angry about the horse.** **The interactions log is the mechanism, which is why it is not optional at this tier.** Every file names its own log and says what specifically to record in it, because what accumulates differs per antagonist: Corwen accumulates *tactics*, Ilma accumulates *what she was told*, Ossa accumulates *refusals*, Iselde accumulates *the shape of the party's attempts*. ### 2. They need an exit, decided up front `exit:` is the **designed** resolution — a commitment by the author about what closure looks like. **It is not a constraint on play.** Anyone here can eat a crit. Its job is to stop a nemesis becoming a recurring tax. **Turned is the best of the three** (three of six), and **killing one should cost the party something** — which is why exactly one is written with no good ending. Six turnable antagonists would make the field decorative. ### 3. Not core to the campaign > **If their death would end the plot, they are a front, not an antagonist.** A constraint on how they are *written*. It is why the war Corwen fought is unnamed, why Ossa's maker is off-screen and must stay there, and why Teodor's real culprit is a campaign decision rather than a fixture. **Every one of these six can be removed from a campaign without the campaign noticing.** --- ## Not the hostile tier [`../tables/roles/hostile/`](../tables/roles/hostile/) holds **10 command-tier role templates** — bandit-captain, crime-boss, spymaster, cult-leader — for threats you *drop in*. Those are lean generation scaffolding with no persistence. **These six are the ones who come back.** **Sharing a stat block across the two tiers is fine and intended.** The hostile spec says persistence is *"none — unless they survive and start recurring, at which point promote them."* A nemesis who is mechanically a bandit-captain is that promotion path working. What must not be shared is the **shape**: if a promoted hostile ends up wanting the same thing as one of the six, pick a different axis or do not promote them. Teodor's own scaling leans on the hostile tier explicitly — at high tier he *hires* a `mercenary-captain`, and the captain is the encounter. --- ## Format Same as [`../adventurers/`](../adventurers/) — frontmatter block, then `Voice / Want / Tell` first because it is the whole point, then Actions, Features, Equipment, `Knows` with visibility tags, Scaling, Provenance. **Three fields only this tier has** — `grudge`, `exit`, `escalation` — plus `grudge_seed` and the `attaches_to` / `fallback` pair, all three added during the build: ```yaml grudge: null # FILL AT CAMPAIGN CREATION — one line, party-relative grudge_seed: "..." # the character-level fact it hangs on. NOT a substitute. exit: resolved | turned | killed escalation: | # what changes on appearance 2, 3, 4+ attaches_to: [] | [party] # added 2026-08-24 — see below fallback: "..." # what to do if that PC is not at the table ``` ### Why `attaches_to` and `fallback` were added **Not in the original spec, and the build needed them.** A party is 3-4 of 14 characters. Four of these six hang off a specific PC's hook, so without a declared dependency the DM discovers at campaign creation that two of their six nemeses are unusable. **Every PC-dependent antagonist carries a fallback that keeps them usable anyway.** `attaches_to: [party]` means no dependency at all — Ilma and Teodor are portable to any composition. ⚠️ **Zephyr carries two of the six** — Maelis via the golem-body, Ossa via the lycanthropy. A party without him drops to four usable antagonists before fallbacks. Both fallbacks are written and both hold, but **check this first** if Zephyr is not at the table, because it is the one composition that thins the bench noticeably. ### Level 3, and asymmetric scaling **All six are anchored at ~3**, because levels 1-3 is where the first playable campaign actually runs, and a nemesis you cannot use until session twelve is a nemesis nobody writes in. **`## Scaling` is weighted hard toward scaling UP** — unlike the adventurers, whose scaling is roughly symmetric. A nemesis met at 2 and still recurring at 9 has to survive a 4x power shift, and surviving it is the section's whole job. **Note what the high-tier entries have in common:** almost none of them are about levels. Ilma gains jurisdiction, Teodor gains a purse, Iselde gains an audience, Ossa gains a population. **The correct tier-3 version of most of these people is a situation the party cannot fight** — which is the honest answer to the scaling problem rather than a stat-block arms race. --- ## Where the interactions logs live **Not here.** These files are **template layer** — the same distinction as `characters/` versus `LIVE//builds/`. ``` npcs/antagonists/.md template (this directory) LIVE//canon/antagonists/-interactions.md instance (created at t=0) ``` **A sibling of `canon/locations/`, not inside it** — antagonists travel, and are not bound to the place they were first met the way a captured NPC is. Activation is a campaign-creation step: pick which of the six are live, fill each `grudge:` against the actual party, and create the logs. See [`../../docs/campaign-start.md`](../../docs/campaign-start.md) §5b and the t=0 manifest in §8. --- ## Provenance Authored 2026-08-24 against the brief in `HANDOFF-v1.md` §4 Task 1 and `npc-containers.md` §8. **Four of six are seeded from character-level facts in the D&D Beyond JSON audit** rather than invented — Jornis's discarded name *Zytharin*, Zephyr's recently-contracted wereboar lycanthropy, Zephyr and Llewellyn's shared golem-body, and Deerx's flaw verbatim. Two are party-wide and portable. **The boundary that kept this out of front territory:** > **Character-level facts are fair game. World-level facts are fronts.** Jornis is Fey and hunted regardless of which campaign runs — that is true about a *person*. *"The Sundered Choir wants to wake the thing under the barrow"* is true about a *world*, and writing it here would be authoring a front by accident. Every hook used passes that test; every world detail is left deliberately unnamed for the campaign to fill.