agentleFS
alfredxwalfredxw/denova/skills/novel-lite/SKILL.md

novel-lite

Skill990 starsChanged 4 months agoApache-2.0View on GitHub

Fast continuation, exploratory drafts, and low-latency prose generation performed directly by the main Agent without review or revision subprocesses.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 7 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this Skill for fast prose generation by the Writing Agent.

    line 12
  2. Distinctive:

    Do not assume the task is the next chapter unless the user explicitly asks for it.

    line 17
  3. Distinctive:

    When the user requests multiple chapters, do only a lightweight internal breakdown: determine the overall direction and chapter boundaries, …

    line 19
  4. Distinctive:

    If the user asks only for a passage, exploratory draft, or example in chat, output the prose directly and do not write workspace files.

    line 28
  5. Very specific:

    When the user asks to create or update workspace prose, use write for a new file or complete rewrite and edit for localized replacement.

    line 29
  6. Very specific:

    After user-visible file changes, use read to verify key excerpts from the final file.

    line 31
  7. Very specific:

    Treat setting/progress.md only as a summary.

    line 33
  8. Distinctive:

    Do not start reviewer, fixer, task, General SubAgent, or configured subagent workflows.

    line 38
  9. Distinctive:

    Use workspace context, selected text, lore references, and style rules only when relevant to this turn's writing scope.

    line 44
  10. Distinctive:

    Add a brief explanation only when the user asks for one or an important constraint cannot be satisfied.

    line 49

≈ “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

novel-lite

Use this Skill for fast prose generation by the Writing Agent.

Determine the writing scope

  • Infer the scope from the user's actual request, such as continuing a passage, writing a scene, one chapter, three chapters, a story arc, or a custom target.
  • Do not assume the task is the next chapter unless the user explicitly asks for it.
  • There is no writing_scope field. The user message is the sole source for scope, goal, constraints, and output form.
  • When the user requests multiple chapters, do only a lightweight internal breakdown: determine the overall direction and chapter boundaries, then write at the requested scale.

Workflow

main Agent -> final output

Tool requirements

  • When continuity matters, first use read for the relevant workspace files, such as CREATOR.md, setting/outline.md, setting/progress.md, setting/character-states.md, the current chapter-group plan, and recent chapters. For lore, use query_lore_items with exact ids/names or search filters and detail=full to load complete bodies. Follow next_offset for additional pages; the default index only returns briefs.
  • If the user asks only for a passage, exploratory draft, or example in chat, output the prose directly and do not write workspace files.
  • When the user asks to create or update workspace prose, use write for a new file or complete rewrite and edit for localized replacement. old_string must match the current file exactly and uniquely and must not contain line-number prefixes returned by read.
  • Inspect every write or edit result. If it contains [tool error], invalid JSON arguments, string not found, a non-unique match, a path error, or a truncation notice, do not claim completion. Reread the target, correct the arguments, and retry, or clearly report that the write did not succeed.
  • After user-visible file changes, use read to verify key excerpts from the final file. If verification is impossible, say so in the final response.
  • After writing a complete chapter or materially changing its plot, finish the prose self-review and final revision, then update setting/progress.md and setting/character-states.md in the same turn. Chapter status in the UI does not affect synchronization. Pure typo, punctuation, or wording edits that do not change narrative facts need no state update.
  • Treat setting/progress.md only as a summary. Determine the next chapter from actual chapter paths and non-empty prose, and correct progress in the same turn when they conflict.

Rules

  • The main Agent produces the final result directly.
  • Do not start reviewer, fixer, task, General SubAgent, or configured subagent workflows.
  • Before drafting, continuing, rewriting, or polishing narrative prose, load novel-craft with the skill tool unless it is already loaded in this turn. Follow its scenario routing and read only the references it names for the current task; for a chapter or scene, start with scene-and-payoff.md; for serial continuity, also read the 记忆与衔接 section of long-form.md. Do not load every reference. If the skill tool is unavailable, continue with the rules here. Skip this for non-narrative edits such as typos, punctuation, formatting, or metadata. Loading novel-craft is a Skill read, not a review subprocess.
  • A lightweight internal check for continuity, user requirements, and obvious prose issues is allowed, but do not expose the review process.
  • For prose, silently remove obvious AI patterns: summary or realization endings, emotion labels, identical reactions across characters, and paragraphs of uniform length and information density. Preserve intentional style and do not add new events just to sound human.
  • Classify fact changes before committing them to continuity. Incidental unnamed details need no record. New named characters, locations, items, relationships, promises, new uses of established abilities, and new hard numbers should be written and then recorded in the existing chapter-group plan, progress, character state, or lore as appropriate. Changes to established rules, core abilities, protagonist identity, major-character death, timeline contradictions, or confirmed outline promises require author confirmation first. Do not create a separate fact registry.
  • Preserve user control. Do not over-plan, over-explain, or turn the requested draft into a different story.
  • Use workspace context, selected text, lore references, and style rules only when relevant to this turn's writing scope.

Output

  • Return the creative result the user requested directly.
  • Add a brief explanation only when the user asks for one or an important constraint cannot be satisfied.

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.