Series Orchestrator

Series Orchestrator — Multi-Book Fiction Pipeline Management

Orchestrates long-form fiction across 2+ books/sagas. Bridges single-book narrative craft with franchise-level systems: character arcs spanning volumes, power-progression tracking, faction relationship matrices, and toolchain/skills assembly.

When to Use

Trigger when:

• A project spans multiple books/sagas/volumes
• You need to track character/power/faction state across books
• A power-magic system has explicit escalation rules (star ratings, cultivation ranks, etc.)
• You need to select/install external skills for a writing pipeline
• An Obsidian vault is needed for story bible/references/workspace
• Book-level plot threads need foreshadowing hooks in earlier books

Skills Audit Process

When assembling a writing pipeline, use this systematic approach:

Step 1: Query the Live Hub

hermes skills browse                          # see available categories
hermes skills search "<multi-word query>"     # use phrases, NOT single words
hermes skills inspect "<skill name>"          # preview content before installing

Pattern: Single-word queries often return zero results. Always use multi-word queries:

• ❌ search "writing" → empty
• ✅ search "creative writing" → 12 results
• ✅ search "obsidian" → 25 results

Step 2: External Verification

After finding candidates in the hub, verify via web:

• Check GitHub stars, README quality, actual content vs description
• Look for author reputation and maintenance activity
• Cross-reference with community discussions (Reddit, dev.to, LinkedIn)
• Note installation commands (usually hermes skills install <id> or npx skills add <repo>)

Step 3: Category Mapping

Map findings to these categories:

Step 4: Install & Evaluate

Install candidate skills, then use them for one real task before deciding to keep. Document which ones proved useful and which didn't.

Priority Tiers for Writing Skills

Based on the Eldrath series audit, priority ordering:

P1 — Critical

• blader/humanizer — Removes 33 AI writing patterns; voice calibration. Install FIRST for any publication-targeted project.
• writing-claw — Narrative OS with character registry, gap-based tension theory, series-universe hierarchy. Critical for multi-book consistency.

P2 — High

• danjdewhurst/story-skills — Story bible format, faction matrices, artifact tracking, chapter/state progress bar. Built for sagas.
• haowjy/creative-writing-skills — Three sub-skills: Story Architecture, Prose Quality, Character Depth. Excellent line-editing guidance.
• the-storytellers-workbench — Craft-level diagnosis engine: flat scenes, voice drift, pacing, character interiority.

P3 — Medium

• modoojunko/awesome-novel-skill — Chinese-developed, fantasy/wuxia templates. Worldbuilding module may complement our approach.
• Base obsidian skill — Foundation for vault management. Novel-specific workflow built custom.

Custom Skill Builds (Don't Exist Yet)

For series projects, these typically need to be built custom because they combine domain knowledge:

Premise Update / Ledger Re-Sync (mid-project)

When the user drops a plot, character, or world update (from Discord, a conversation, a brainstorm, a voice memo), propagate to ALL ledger files before drafting anything new. Order matters — downstream files depend on upstream:

1. Read every ledger file — plot-ledger, character-sheet, canon, foreshadow-bank, worldbuilding, concept. Identify what changed vs. what the user said.
2. character-sheet.md — new/renamed characters, species, roles, combat styles. Update sliders if arc changed.
3. canon.md — names, places, rules, timeline. New facts go here first; never let a fact appear in a chapter before it's in canon.
4. plot-ledger.md — re-map beats. If a single beat now covers multiple distinct events (e.g. "three dungeons" was one row), split into separate scene-level rows with their own scene IDs. Every new event gets a beat row.
5. foreshadow-bank.md — re-index:
   • New beats → new plant rows (especially trust-building scenes that set up later betrayal/payoff).
   • Changed chapter assignments → fix plant/payoff chapter columns.
   • Deleted beats → remove or mark red-herring.
   • Verify: every payoff row has a matching plant; no plant points to a chapter that no longer exists.
6. worldbuilding.md — new species, factions, magic rules, societal norms.
7. concept.md — update logline/premise notes if the story engine changed.
8. Diff report to user — table of files changed + what changed. Flag any [TBD] fields still open.

Pitfall: updating one ledger file and forgetting the others. Canon says "three dungeons" but plot-ledger still says "first dungeon." Foreshadow bank points to ch05 for a rescue that now happens in ch04. Always propagate to ALL files in one pass.

Cross-Book Continuity Checklist

Before starting Book N, verify:

• All cliffhanger hooks from Book N-1 addressed or explicitly deferred
• Protagonist's power level is consistent with where they ended Book N-1
• New characters introduced have proper backstory established
• Faction relationships evolved logically (no sudden reversals without cause)
• Recurring locations/items have consistent descriptions
• Foreshadowing from Book N-1 pays off OR is retconned with justification
• Supporting character aging/timeline remains plausible
• Power scaling hasn't inflated beyond the system's internal logic

Power System Tracking Template

For explicit power-scaling systems (star ratings, cultivation ranks, etc.):

Maintain a power-progression.yaml per main character:

character: Noctis
book_1_entry:
  rating: 0  # unknown/low
  abilities: ["basic aura sensing"]
  limitations: ["incomplete training", "fear of full power expression"]
book_1_climax:
  rating: 2
  abilities: ["aura manipulation", "elemental affinity"]
book_2_entry:
  rating: 3
  # ...

Update at end of each drafting pass. Never let a character jump more than 1-2 stars per book without a compelling arc.

Obsidian Vault Pattern (Series-Scale)

The vault IS the project root — not a parallel directory. One vault hosts ALL books + shared world lore. Structure:

{series-name}/                         ← vault root (e.g. the-eternal-frontier/)
├── 00-HOME.md                         ← Root MOC — single entry point, wikilinks to all
├── 01-World/                          ← SHARED across all books
│   ├── worldbuilding.md               ← pillars, magic laws, societal implications
│   ├── canon.md                       ← truth ledger (names, rules, timeline)
│   ├── factions/                      ← one file per faction
│   ├── species/                       ← one file per species
│   ├── magic-system/                  ← one file per magic subsystem
│   └── locations/                     ← one file per location
├── 02-Characters/                     ← SHARED — one file per character (spans all books)
│   ├── {protagonist}.md               ← full template (character-builder skill)
│   ├── {companions}.md
│   └── {antagonists}.md
├── 03-Books/                          ← Per-book ISOLATED working files
│   ├── book1-{slug}/
│   │   ├── concept.md
│   │   ├── plot-ledger.md
│   │   ├── foreshadow-bank.md
│   │   ├── voice-profile.md
│   │   ├── manuscript.yaml
│   │   ├── chapters/
│   │   ├── outlines/
│   │   └── reviews/
│   ├── book2-{slug}/
│   └── ...
├── 04-Series/                         ← Cross-book tracking
│   ├── series-overview.md             ← N-book arc summary + cross-book threads
│   ├── power-progression.md           ← protagonist star-rating per book
│   └── thread-tracker.md              ← open/resolved threads across books
└── 05-Resources/                      ← Meta / pipeline / craft references
    └── style-guide.md                 ← voice rules, banned words, tone defaults

Design principles

1. World and Characters are SHARED — they live outside any single book so Book 2–5 reference the same canon. Character sheets track per-book arc state inside the file.
2. Books are ISOLATED — each book has its own ledger, foreshadow, chapters, reviews. No cross-contamination.
3. 00-HOME.md is the MOC — every file is reachable via wikilinks from HOME. No orphans.
4. Wikilinks everywhere — [[Noctis]], [[Saora]], [[Star Ratings]] connect the graph.
5. Folder numbering — 00-05 prefix gives sort order and signals hierarchy at a glance.
6. Series tracking (04-) — power progression, thread tracker, and overview live here because they span books.

Build order for a new series vault

1. Create folder skeleton (all dirs)
2. Write 00-HOME.md with wikilinks (even to files that don't exist yet — Obsidian shows them as unresolved)
3. Move/copy existing flat files into correct positions
4. Create stubs for factions, species, locations (headers + known facts + open questions)
5. Create character files using character-builder skill template
6. Create series tracking files (overview, power-progression, thread-tracker)
7. Create style-guide in 05-Resources
8. Update titles/headers in moved files to reflect new series/book names
9. Delete old flat directory (with user confirmation)
10. Update memory with canonical vault path

Consolidating duplicate trees

Projects often accumulate two manuscript trees — e.g. an early Obsidian vault with an old plot structure AND a pipeline manuscript/{slug}/ directory with the current structure. When this happens:

1. Identify which tree is canonical — the one matching the user's latest premise/plot updates.
2. Merge any unique content from the stale tree (voice samples, reviews, outlines) into the canonical tree.
3. Delete the stale tree entirely. Two trees = guaranteed canon drift.
4. Update memory with the surviving canonical path so future sessions don't re-discover the old one.
5. Report to user what was kept, what was merged, what was deleted.

Never draft against a tree without confirming it's the only one.

See references/skills-audit-guide.md for the complete audit methodology and source URLs. See references/build-order-discipline.md for the non-negotiable pre-draft sequence (characters → world → voice → draft). See references/character-complexity-scaling.md for character sheet complexity tiers, working-document style rules, and the three-layer review system (doc-review / narrative-revisor / humanizer).

Pitfalls

• Hub search requires multi-word queries. hermes skills search "writing" returns empty. Use hermes skills search "creative writing" or "fiction storytelling".
• Description ≠ substance. Always inspect a skill and read the SKILL.md body before installing. Many descriptions are thin.
• External GitHub repos may be unmaintained. Check last commit date, issue open rate, and actual file count. Some popular repos have sparse codebases.
• Chinese-language skills may have English READMEs but Chinese code/templates. Verify language coverage matches your project language.
• Don't over-install. Test one skill per category. Keep only what you actually use. Bloat slows skill loading.
• Duplicate manuscript trees. User may have an old Obsidian vault AND a pipeline tree for the same project. Always check for multiple trees before drafting. Consolidate into ONE canonical path, delete the stale one, update memory.
• Premise updates not propagated to all ledgers. When the user drops new info, ALL files must re-sync in one pass (character-sheet → canon → plot-ledger → foreshadow-bank → worldbuilding → concept). Updating one file and forgetting the others creates silent canon drift.
• Thin-draft trap (cross-ref: narrative skill). When drafting chapters, hitting plot beats without hitting word count targets. Chapter outlines typically specify ~7,500 words each. If chapters land at <2,000 words, you're writing beat summaries instead of full scenes. Expand with sensory detail, dialogue, interiority, action. Run word count check before marking chapters done. The outline is the map, not the territory.