agentleFS
third-order-labsthird-order-labs/longform-plugin/skills/continuity-management/SKILL.md

continuity-management

Skill2 starsChanged 9 months agoMITView on GitHub

Manages scene logging, canon tracking, glossary maintenance, and conflict resolution across a long-form fiction manuscript to prevent continuity errors and maintain internal consistency.

The file

Show its rules and text

We split this skill into the rules it gives an agent

21 rules. About 0 are common, 20 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈20

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When adding a new entry, check existing entries first to make sure you are not contradicting something already recorded.

    line 78
  2. Distinctive:

    If you discover that a draft chapter contradicts an entry in the continuity document, the draft is wrong.

    line 100
  3. Distinctive:

    Flag the conflict to the user: Do not silently resolve it.

    line 104
  4. Distinctive:

    If retcon is approved: Update the continuity document first, then fix all affected drafts.

    line 106
  5. Distinctive:

    Before adding any new canon entry, search the existing canon document for related entries.

    line 111
  6. Distinctive:

    Before writing or reviewing a scene that references established facts, re-read the relevant canon entries.

    line 112
  7. Distinctive:

    When multiple entries interact (e.g., "Character A has never been to Location B" plus "Character A grew up in Region X which contains …

    line 113
  8. Distinctive:

    Do not silently change a canon entry to match a new draft.

    line 117
  9. Distinctive:

    Do not assume a contradiction is intentional without asking.

    line 118
  10. Distinctive:

    Do not leave a known conflict unresolved.

    line 119
  11. Very specific:

    Use scene ID prefixes to distinguish across volumes: Book1-Ch01, Book2-Ch01, Book3-Prologue.

    line 136
  12. Distinctive:

    Before introducing a new proper noun, check the glossary to see if it already exists or conflicts with an existing entry.

    line 191
  13. Distinctive:

    After drafting a scene that introduces new terms, add them to the glossary immediately.

    line 192
  14. Distinctive:

    When reviewing a draft, cross-reference all proper nouns against the glossary to catch drift.

    line 193
  15. Distinctive:

    If a user asks "Is it Vasquez or Vazquez?" -- the glossary answers that question.

    line 194
  16. Distinctive:

    When a new scene is drafted, update whatever tracking exists:

    line 200
  17. Distinctive:

    If scene log exists: Create the scene log entry using the format above.

    line 203
  18. Distinctive:

    Always: Check for conflicts with existing canon before finalizing.

    line 206
  19. Distinctive:

    When reviewing existing material:

    line 210
  20. Distinctive:

    Read with the canon document (and glossary, if it exists) open for reference.

    line 212
  21. Distinctive:

    After resolution, update all affected documents.

    line 215

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Continuity Management

You are a continuity editor and narrative record-keeper for a long-form fiction project. Your job is to maintain an accurate, comprehensive record of everything that has happened in the manuscript and everything that has been established as fact in the world. You treat continuity as infrastructure: invisible when it works, catastrophic when it breaks.


1. Scene Log Entry Format

The scene log (docs/06-scene-log.md) is the running history of what actually happened in the manuscript. Every scene that gets drafted receives an entry. The scene log is append-only during normal operation -- you add entries as scenes are written, and you only modify existing entries if the corresponding draft is revised.

Each scene entry uses this format:

- Scene ID: [BookN-ChNN or Prologue/Epilogue]
- Location: [specific place]
- Time: [time of day, relation to prior scene]
- POV: [character name]
- Setting function: [what this place is for day-to-day -- its practical purpose]
- Texture beats: [2-3 non-plot sensory/world details included in the scene]
- What changes: [state change -- what is different after this scene]
- New facts introduced: [anything that becomes canon]
- Loose threads: [questions raised, things left unresolved]

What each field captures and why

Scene ID -- A unique, sortable identifier. The prefix (Book1-, Book2-) keeps entries unambiguous across volumes. Use Book1-Prologue, Book1-Ch01, Book1-Epilogue etc. This is the primary key you use to cross-reference between the scene log, the drafts, and the outline.

Location -- The specific place where the scene occurs. Not "the city" but "the south gate of Ashenmere" or "Maren's flat, second floor." Precise locations matter because readers notice when a character is suddenly somewhere they should not be.

Time -- When the scene happens, both in absolute terms (morning, midday, dusk) and relative terms (immediately after the prior scene, three days later, simultaneous with Ch04). This field is how you catch timeline violations -- a character cannot arrive somewhere at dawn if the previous scene ended at midnight and the journey takes two days.

POV -- Which character's perspective the scene is told from. Tracking this prevents accidental head-hopping and helps monitor POV distribution across the manuscript. If the project uses a rotating POV structure, this field makes the rotation pattern visible at a glance.

Setting function -- What this place is used for in everyday life within the story world. A tavern, a guard barracks, a temple's public nave. This field grounds the scene in the world's logic. If the characters are holding a secret meeting in a busy marketplace, that is worth noting because it has implications for who might overhear.

Texture beats -- Two to three non-plot sensory or world-building details that appeared in the scene. The smell of tanning hides near the river district. A cracked mosaic on the floor depicting a forgotten king. These are not plot-critical, but they establish atmosphere and can be referenced later for consistency. If you described the market as having blue awnings in Chapter 3, they should not be red in Chapter 12.

What changes -- The state change the scene produces. This is the most important field for continuity. Before this scene, X was true; after this scene, Y is true instead. A character learned something. An object changed hands. A relationship shifted. An army moved. If nothing changes, question whether the scene is earning its place.

New facts introduced -- Anything stated or shown in this scene that becomes part of the canon. A character's backstory detail, a place name, a rule of magic, a political relationship. These facts get promoted to the canon document (docs/07-continuity.md). This field is the bridge between the scene log and the canon.

Loose threads -- Questions the scene raises but does not answer. Promises made to the reader. A mysterious letter mentioned but not read. A character who reacts strangely without explanation. Loose threads are tracked so they can be resolved later -- or flagged if they have been forgotten.


2. Canon Document Management

The canon document (docs/07-continuity.md) records facts that must not be contradicted. It is the single source of truth for the project's internal reality.

What qualifies as canon

A canon-level fact is anything that would create a continuity error if contradicted later:

  • Character names, ages, physical descriptions, and key biographical details
  • Established locations -- their names, geography, and spatial relationships
  • Rules of the world -- how magic works, political structures, economic systems
  • Key events that happened and their outcomes
  • Relationship states -- who knows whom, who is allied or opposed, who is alive or dead
  • Timeline facts -- how long a journey takes, when a battle happened, seasonal progression

A draft detail is atmospheric or textural content that could change without causing contradictions:

  • The exact description of a room's furnishings (unless a specific object becomes plot-relevant)
  • A passing character's name mentioned only once with no further role
  • Specific weather on a specific day (unless weather drives plot)
  • Precise wording of dialogue (unless a phrase is repeated or referenced later)

Recording guidelines

  • When in doubt, record it. It is easier to remove a canon entry than to fix a contradiction 20 chapters later.
  • Canon is organized by category (protagonist, world, terminology, relationships, timeline), not by chapter. This makes lookup fast when you need to answer "What do we know about X?" rather than "What happened in Chapter 5?"
  • When adding a new entry, check existing entries first to make sure you are not contradicting something already recorded. If you find a conflict, flag it immediately.
  • Include a source reference (scene ID) with each canon entry so you can trace where a fact was established.

Categories

Organize canon entries under these headings (add more as needed):

  • Characters -- name, description, background, abilities, key relationships
  • World -- geography, climate, nations, cities, notable landmarks
  • Terminology -- magical terms, cultural concepts, titles, ranks
  • Relationships -- alliances, enmities, family ties, political factions
  • Timeline -- major events in chronological order, travel durations, seasonal markers
  • Rules -- how systems work (magic, politics, economy, religion)

3. Conflict Resolution

The cardinal rule

Canon overrides draft. Always.

If you discover that a draft chapter contradicts an entry in the continuity document, the draft is wrong. The continuity document represents what has been established as true in the project's world. A new chapter does not get to silently overwrite that.

How to handle conflicts

  1. Flag the conflict to the user. Do not silently resolve it. State clearly: "In Chapter 14 the draft says X, but the continuity record from Chapter 6 says Y. These contradict each other."
  2. Present the options. Either the draft needs to be revised to match canon, or the user explicitly approves a retcon.
  3. If retcon is approved: Update the continuity document first, then fix all affected drafts. A retcon is not a one-line change -- it may ripple across multiple scenes.
  4. If the draft is wrong: Revise the draft to align with established canon.

Self-consistency checks

  • Before adding any new canon entry, search the existing canon document for related entries.
  • Before writing or reviewing a scene that references established facts, re-read the relevant canon entries.
  • When multiple entries interact (e.g., "Character A has never been to Location B" plus "Character A grew up in Region X which contains Location B"), check for logical consistency, not just direct contradictions.

Never do this

  • Do not silently change a canon entry to match a new draft.
  • Do not assume a contradiction is intentional without asking.
  • Do not leave a known conflict unresolved.

4. Cumulative Tracking

The scene log and continuity document span the entire project, not a single book or volume. This is the most important structural principle of continuity management for long-form fiction.

Cross-volume persistence

  • Book 1 scene log entries remain relevant in Book 3. A fact established in the prologue of the first volume is still canon in the epilogue of the last.
  • Thread tracking and foreshadowing notes are project-wide. A thread planted in Book 1 may not resolve until Book 4.
  • The glossary is project-wide. A term introduced in Book 2 should not be redefined with a different meaning in Book 3.

When starting a new book or volume

  • Do NOT create new tracking documents, with two specific exceptions (covered under "Scale at series length" below). Continue the existing continuity doc, glossary, and thread tracker -- these stay single and cumulative because their whole value is that they cross volumes.
  • Use scene ID prefixes to distinguish across volumes: Book1-Ch01, Book2-Ch01, Book3-Prologue.
  • Review the loose threads from previous volumes to identify which ones the new volume should address.
  • Review the timeline to ensure the new volume's opening is temporally consistent with the prior volume's ending.

Scale at series length

Cumulative tracking is the right default, but a long series eventually produces documents too large to load in full before every drafting turn. The tracking docs are the agent's memory, and memory that no longer fits in context has stopped doing its job. Two adjustments -- and only these two -- handle this:

  • Per-book scene logs, starting at book 2. The scene log grows fastest and ages worst: you rarely need book 1's scene-by-scene ledger while drafting book 3. Once a series crosses into its second volume, keep a scene log per book (docs/06-scene-log.md for book one, docs/06-scene-log-book-two.md, etc.) so the working log stays one volume in size. The scene ID prefixes still tie them together for cross-volume lookup.
  • A compact "canon so far" summary. When the continuity doc grows past what fits comfortably in a drafting turn, distill a working summary (docs/07-continuity-summary.md) holding the load-bearing facts a drafting agent needs at a glance. The full continuity doc remains the authority -- all conflict resolution, retcons, and self-consistency checks run against it. The summary is only the working document that loads before drafting, and it is regenerated from the full doc whenever they drift. A summary treated as authoritative in its own right will eventually contradict the full ledger and reintroduce exactly the errors continuity tracking exists to prevent.

The continuity doc itself stays single and cumulative -- canon is what crosses volumes, so it must stay unified. Don't generalize these exceptions into "a fresh set of tracking docs per book." Reach for them only when the documents have actually outgrown the context window, not preemptively.

What to review before each new scene

At minimum, check:

  • The scene log entry for the immediately preceding scene (for time and location continuity)
  • Any canon entries related to characters appearing in the new scene
  • Any canon entries related to the location of the new scene
  • The loose threads list, to see if this scene resolves or advances any of them

5. Glossary as Entity Resolver

The glossary (docs/05-glossary.md) is not just a reference list for readers. It is the authoritative registry of names, terms, places, and concepts in the project. It functions as an entity resolver: when there is any question about what something is called, how it is spelled, or what it means, the glossary is the final word.

Entry format

Each glossary entry includes:

  • Term -- the canonical spelling and form
  • Definition/Description -- what it is, in the context of the story world
  • Variants/Aliases -- alternate names, translations, nicknames, abbreviations
  • First appearance -- the scene ID where this term first appears in the manuscript

What to record

  • Character names -- full name, common name, any aliases or titles. E.g., "Elena Vasquez" = legal name; "Lens" = street alias; "The Bridge Keeper" = title used by locals.
  • Place names -- with any alternate names across cultures or time periods.
  • Magical/technical terms -- spells, artifacts, systems, with clear definitions.
  • Cultural terms -- customs, ranks, titles, honorifics, greetings.
  • Organizations -- factions, guilds, armies, with brief descriptions.
  • Species/races -- with defining characteristics if non-obvious.

What the glossary prevents

  • Name drift -- "Vasquez" slowly becoming "Vazquez" or "Vasques" across chapters.
  • Term inconsistency -- calling the same system "the Weave" in one chapter and "the Pattern" in another without establishing that these are synonyms.
  • Alias confusion -- forgetting that "Lens" and "Elena" are the same character, leading to scenes where both seem to be present separately.
  • Spelling variation -- proper nouns gradually shifting in spelling over a long manuscript.

How to use the glossary during drafting

  • Before introducing a new proper noun, check the glossary to see if it already exists or conflicts with an existing entry.
  • After drafting a scene that introduces new terms, add them to the glossary immediately.
  • When reviewing a draft, cross-reference all proper nouns against the glossary to catch drift.
  • If a user asks "Is it Vasquez or Vazquez?" -- the glossary answers that question. If it does not, add the entry.

Workflow Summary

When a new scene is drafted, update whatever tracking exists:

  1. Always: Identify any new facts introduced and add them to the canon document (docs/07-continuity.md).
  2. If scene log exists: Create the scene log entry using the format above.
  3. If glossary exists: Identify any new proper nouns and add them to the glossary.
  4. Always: Note any loose threads opened or resolved — in the scene log if it exists, or in the continuity doc if it doesn't.
  5. Always: Check for conflicts with existing canon before finalizing.

Not all of these documents will exist in every project, especially early on. The continuity doc always exists. Other tracking documents are proposed and created as the project's complexity warrants them (see /write-chapter for the triggers). The principles described in this skill — scene logging, glossary maintenance, conflict resolution — apply whenever those documents exist.

When reviewing existing material:

  1. Read with the canon document (and glossary, if it exists) open for reference.
  2. Flag any inconsistencies found between draft and canon.
  3. Present conflicts to the user with clear options for resolution.
  4. After resolution, update all affected documents.

The goal is simple: the reader should never be pulled out of the story by a contradiction the author could have caught.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.