Desk-Reject Risk (rt-desk-reject-risk)

Desk rejection is the modal outcome at a strong venue, it arrives before any referee reads the paper, and it costs weeks for reasons that are almost always documented in advance. This repository already carries those reasons — 456 of 743 indexed venues (61%) publish an explicit desk-reject section in their pack — but nothing turned them into a check a paper could be scored against.

rt-submission-readiness answers "is the manuscript mechanically complete?". This answers the different and harder question: "given this specific venue's stated triggers, what will get this desk-rejected, and what is the cheapest fix?"

When to trigger

• A target venue is chosen and the paper is near-final.
• An author asks "will this survive the editor?" or "why did the last one get desked?"
• Immediately after rt-venue-reframe, to score the reframed draft against its new home.
• Before a resubmission up the ladder, where the previous desk reject is evidence.

Method

Step 1 — locate the venue's triggers. From venue-index.tsv, take the row's pack_dir (depth) or profile_path (breadth) and read:

• the ## Common desk-reject triggers section (or the pack's Desk-reject and triage / decline-without-review variant) in its *-submission, *-review-process and *-topic-selection skills;
• the scope and format constraints in resources/official-source-map.md, read live — length caps, abstract style, anonymity rules and fee/policy gates are volatile and are a large share of real desk rejects.

If the venue publishes no explicit trigger list (about 4 in 10), say so, and fall back to the venue's fit and house-style skills plus the generic gate in Step 3 — do not invent venue-specific triggers.

Step 2 — score each trigger. For every documented trigger, judge the manuscript against it and assign:

Cite the evidence in the manuscript for each judgement (a section, a table, a claim), not a general impression. A risk with no located evidence is a hypothesis — mark it as one, or drop it.

Step 3 — the universal gate (applies at every venue, on top of the venue's own):

1. Scope miss — the contribution is not the kind of contribution this venue publishes. This is the single largest desk cause and the least fixable.
2. Framing miss — right result, wrong "why it matters" for this audience.
3. Format/anonymity breach — length, abstract, blinding, reference style, ORCID, supplementary rules. Mechanical and always worth fixing.
4. Evidence baseline — a robustness or design check the venue treats as table stakes and the paper omits.
5. Policy — data-and-code availability, ethics/IRB, preregistration, conflicts, prior dissemination or dual-submission rules.

Step 4 — return a cost-ranked fix list. Sort by (severity × 1/effort). Distinguish fixes that are edits from fixes that need new analysis; say explicitly when the honest conclusion is "this venue is wrong for this paper" and route to rt-journal-match.

Hard rules

1. Venue-conditional, never generic. Every Fatal/Likely must trace to a trigger the venue's own pack or source map states. Generic advice belongs in Step 3 and must be labelled as such.
2. Volatile constraints are read at scoring time from resources/official-source-map.md — never quoted from memory.
3. No fabricated triggers. If the pack does not document a trigger list, say so.
4. Do not present this as a prediction of the editor's decision. It is a documented-risk audit; there is no acceptance-probability number to give, and inventing one would be false precision.
5. Defer scope/fit judgement to the venue's *-topic-selection skill.

Output format

【Venue】<display name>  (pack: <pack_dir | profile_path>)
【Trigger source】<n> documented triggers read from <files>  |  none documented
【Source-map read】<date>
【Fatal】  trigger — evidence in the manuscript — fix (edit | new analysis)
【Likely】 …
【Watch】  …
【Clear】  triggers the paper demonstrably passes
【Cost-ranked fixes】1..n, each: severity · effort · what to change
【Verdict】submit as-is / fix first (n items) / wrong venue → rt-journal-match

Anti-patterns

• Producing a percentage "acceptance chance". The inputs do not support it.
• Listing every trigger as Watch — a report with no ranking is not a report.
• Scoring against a venue's reputation instead of its documented triggers.
• Running this instead of rt-submission-readiness; they check different things and the mechanical bar comes first.
• Treating a Clear on all triggers as an acceptance signal. Surviving the desk means the paper reaches referees, nothing more.

Next: rt-simulated-referee once the desk gate is clear, then rt-replication-package for the deposit.