manuscript-release-check
Skill4 starsChanged 31 days agoMITView on GitHub
Use when checking manuscript finalization, source-to-PDF identity, editor synchronization, submission evidence, or preparing a recoverable handoff for another agent. Especially useful after late source edits, stale previews, missing build dependencies, or conflicting completion reports.
In this repository17 other files it gives its agents.
- academic-prose-editingSkill
- academic-writing-auditSkill
- claim-evidence-auditSkill
- experiment-comparison-auditSkill
- figure-caption-auditSkill
- latex-bibliography-auditSkill
- method-definition-auditSkill
- nondefensive-writingSkill
- reference-authenticity-auditSkill
- related-work-synthesisSkill
- research-manuscript-writingSkill
- results-interpretationSkill
- reviewer-comment-triageSkill
- reviewer-response-writingSkill
- revision-integrity-auditSkill
- scientific-story-auditSkill
- word-citation-auditSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 15
Read current artifacts independently of earlier completion summaries.
- Distinctive: line 25
When the release claims verified BibTeX, require a per-entry ledger linking the intended publication, its two independent identity checks, …
- Distinctive: line 29
Use these fields when another agent must continue.
- Distinctive: line 38
Use real identifiers already available or hashes actually calculated.
≈ “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.
Files
Manuscript Release Check
Tie every completion claim to an identifiable artifact and observed evidence. Source correctness, successful compilation, visual inspection, transfer, and submission are distinct states.
Establish the target
Identify the intended manuscript source, author-approved changes, destination, and requested operation. Resolve “final” or “latest” using actual file locations, revisions, or manifests. Preserve existing sources, assets, and recoverable prior versions. This skill grants no authorization to upload, submit, publish, delete, or replace material; perform external actions only within the user's existing instructions. A readiness review remains a review.
Read current artifacts independently of earlier completion summaries. Separate observations from logs, human reports, assumptions, and your own newly performed checks. If only an evidence packet is available, assess that packet without claiming direct access to its underlying files or services.
Follow the artifact chain
- Source: Identify the entry file and needed bibliography, styles, figures, and other dependencies. Check that the intended edits are in the source being built. If a file changes after verification, earlier evidence remains attached to the earlier version.
- Build: Associate a successful compiler result with its exact input and output. A missing style file establishes a dependency blocker, not that the latest edit is wrong or that the remaining source is valid. Use the legitimate dependency or intended build environment when authorized; do not silently substitute a style or revert approved text to obtain a PDF. A Recompile click is only an action record.
- Render: Inspect the intended PDF and record pages or regions checked. A successful build does not establish unclipped figures, correct affiliations, usable cross-references, or final pagination. Partial inspection supports a partial verdict. A preview showing old text cannot verify the new version.
- Transfer: Distinguish a recorded paste or upload from independently verified persisted contents. Confirm the destination identity and intended source or artifact when accessible. A source transfer, successful remote build, and matching exported PDF require separate evidence; none follows merely from an editor action log.
- Submission: Inspect the actual state or receipt when this is within scope. An attachment listed in a draft confirms neither final submission nor public availability. Link any observed receipt to the intended artifact and action. Do not infer completion from a button click or present a future step as already performed.
When the release claims verified BibTeX, require a per-entry ledger linking the intended publication, its two independent identity checks, its actual Google Scholar Cite → BibTeX capture and the matching complete entry in the built source. Report coverage and unresolved items separately from build success.
Leave a recoverable handoff
Use these fields when another agent must continue; omit only genuinely irrelevant fields:
- Current source: location, entry file, revision or observed identifier, and approved changes to preserve.
- Prior recoverable state: preserved source/artifact locations and the relationship to the current version.
- Build state: input identity, environment/dependencies, command or build entry point when known, log/result, and output identity.
- Render coverage: exact output inspected, pages/regions, material findings, and unchecked areas.
- Destination state: project/file or portal identity, recorded transfer activity, verification evidence, attachment identity, and submission/receipt status.
- Resume actions: smallest remaining checks, unresolved blockers, and action boundaries from the user's instructions.
Use real identifiers already available or hashes actually calculated; do not invent hashes to make the handoff appear precise. A timestamp helps locate evidence but alone does not prove two artifacts match. Keep private paths and identifiers within the authorized handoff context.
Prose when drafting or revising
When drafting or revising prose, including audit explanations, read prose editing rules. Preserve scientific meaning, necessary reasoning and valid technical terms. Keep sound passages, do not add unreported properties, and limit edits to the requested scope. The local reference covers concept accuracy, clear sentences, repetition and punctuation, including the required exceptions.
Report and stop
Lead with the supported readiness judgment, followed by confirmed, failed/blocked, and unverified steps. Name what changed and what still needs checking. A missing receipt means completion is unconfirmed, not proof that a remote action never occurred.
Stop when the requested chain has been verified or every remaining gap has a specific recoverable next step. Preserve the author's valid changes while resolving environmental failures. Do not repeatedly rebuild unchanged inputs, erase prior artifacts, widen a bounded review into publication, or claim that source inspection proves deployment or submission.
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
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.

