Manuscript Editorial Router

Decide which critique pass(es) the request needs, then run them. Do not produce a critique yourself — your output is a routing decision plus the dispatched skill(s). If the request already clearly names one pass, skip the routing prose and run it.

STEP 0 — Resolve the project profile

From the current working directory, walk up the parent folders for .manuscript/profile.md.

• Found → read it. Its genre and passes decide which passes are even

applicable (e.g. fair-play only for mystery/crime/thriller).

• Not found → if the manuscript-profile-setup skill is installed (it ships

with the Manuscript Editorial Suite), invoke it to create one, then continue. If it isn't installed (this router can run standalone), route with genre-general defaults and say so. The passes this router dispatches to are separate skills — install the suite for full function.

The six passes

Supporting (not a critique pass): manuscript-canon-ledger — the living, versioned canon store the continuity pass reads before auditing. Route directly to it for ledger maintenance: "update the ledger", "record a canon change", "is this canon", "resolve open question", "seed the ledger from book N".

Routing rules

1. Disambiguate overlapping phrases by object, not verb:

• "would this sell" + a manuscript/chapter → manuscript-publisher-critique;

+ a blurb/description/listing → manuscript-listing-critique.

• "check the plot logic / does it hold up" about clues and the solution →

manuscript-fair-play-audit; about overall craft → publisher-critique.

• "did I get the details right / does this break canon" → continuity-audit.
• "line audit / is this immersive / kill the info-dumping" → prose-immersion-audit

(line-level), vs. a macro "is the voice working overall" → publisher-critique's Voice row. A request to rewrite prose is neither — route it to the author's writing/voice skill.

• "cinematic pass / movie screen audit / blocking / where is everyone / did we

account for everyone / read this like a director / would I see this on screen" → manuscript-cinematic-scene-audit. Key distinction from prose-immersion: prose-immersion asks "is the writing alive?"; cinematic asks "does the scene make spatial, physical, and emotional sense if you filmed it?" They run together on the same material without overlap.

1. Manuscript vs. listing is the first fork. Selling copy or metadata →

manuscript-listing-critique and nothing else. Story content (scene, chapter, outline, manuscript, query) → one of the other five.

1. Gate by genre. Only run manuscript-fair-play-audit if the profile's genre

is a mystery/crime/thriller (or fair-play is in passes). For a memoir, literary novel, or romance with no puzzle, skip it and say why.

1. Order when multiple passes apply (run sequentially, report each separately):
2. manuscript-continuity-audit — catch canon/consistency breaks first; they

cap the craft grade.

1. manuscript-fair-play-audit — if there's a mystery to judge.
2. manuscript-cinematic-scene-audit — spatial/blocking audit; surface

structural scene problems before going to line level.

1. manuscript-prose-immersion-audit — the line-level craft pass on the scenes.
2. manuscript-publisher-critique — the macro craft/acquisitions read, last, so

it can reference the prior findings. manuscript-listing-critique is independent — run it only for listing material.

1. Broad asks default to a full pass. "Review my book / is this ready / full

editorial pass" on a manuscript → continuity → (fair-play if applicable) → cinematic → prose-immersion → publisher, in that order. For a long manuscript, cinematic and prose-immersion can sample representative scenes — say which.

1. Only one obvious pass? Skip the routing explanation and invoke it directly.
2. Genuinely ambiguous and consequential? Ask one short clarifying question

(e.g. "craft read, canon check, mystery-logic check, cinematic/blocking pass, line audit, or listing copy?") rather than guessing.

Running a selected pass

After selecting the pass(es), for each one load that skill's `SKILL.md` and its referenced files, then follow that skill's required output exactly. Do not produce a critique from the router body alone or paraphrase a pass's format — the value is in each skill's own report structure. Run the passes in the order decided above and present each skill's output as its own clearly-labeled section.

Output

Lead with a one-line routing decision — which pass(es), in what order, and why — then load and run them. Example: "This spans continuity and craft → running continuity-audit first, then publisher-critique." Keep the routing note to one or two sentences; the value is in the passes, not the dispatch.