nih-biosketch

name

nih-biosketch

description

Generate content for an NIH Biographical Sketch (Common Form) on SciENcv, including the Products sections (C.1 closely related, C.2 other significant) and the Biographical Sketch Supplement (Personal Statement, Honors, Contributions to Science). Use this skill whenever the user asks to draft, fill, write, or update an NIH biosketch, SciENcv biosketch, biographical sketch common form, or NIH biosketch supplement, or says "/nih-biosketch". Trigger even when the format isn't named explicitly, if the user mentions preparing biosketch content for a specific NIH grant application (R01, R03, R21, U01, P01, K-series, F-series, T-series, etc.).

NIH Biosketch Generator

Produces SciENcv-ready content for the NIH Biographical Sketch (Common Form) and Biographical Sketch Supplement. Every product, honor, and citation in the output must be traceable to a line in the user's CV.

Required inputs

Two inputs are needed before producing any output:

1. The user's CV, as PDF or DOCX.
2. Grant documents that describe the proposed project. At minimum, Specific Aims. Research Strategy is also accepted if provided. More than one grant document is fine; read all of them.

If either input is missing, stop and ask the user to provide it. Do not draft from partial inputs. Do not guess at a CV or a project from prior context.

If multiple files appear to be CVs (e.g., two PDFs both looking like CVs), ask the user which one to use. Do not pick on the user's behalf.

If the same person has supplied separate Specific Aims and Research Strategy documents, use both.

Reading the inputs

The execution environment provides a file-reading skill router and a pdf-reading skill. Use them. For DOCX, use the docx skill or python-docx tooling.

From the CV, extract:

• Publications, with PMID, PMCID, or DOI when listed
• Other products: software, datasets, patents, technologies, websites, talks, preprints, etc.
• Honors and awards (each with year and organization)
• Positions and training (for the Personal Statement)
• Ongoing and completed research projects from roughly the past three years (for the Personal Statement)

From the grant documents, extract:

• The scientific aims and central hypothesis
• The approach and methods, with attention to techniques the user has in their CV
• The disease, system, or scientific domain being targeted
• The user's stated role on the project, if mentioned

The grant content drives which CV products get pulled into C.1 and how the Personal Statement is framed.

Hard rules on grounding

These are absolute. They exist because biosketches are reviewed by NIH staff and study sections, and fabricated content destroys credibility.

• Do not invent products, publications, or citations that are not in the CV. If you cannot find evidence in the CV, do not list it.
• Do not edit, paraphrase, or improve CV content. The CV is the source of record.
• Do not generate citations beyond what is in the CV. Every product and every in-text reference must trace to a line in the CV.
• Do not invent impact claims (citation counts, awards, adoption stats, media coverage) that are not in the CV.
• If the CV is thin in an area the grant calls for, say so in the output. Do not paper over gaps with invented relevance.
• If the user later asks to add a paper, software, or honor that is not in the CV, refuse and tell them to update the CV first.

Output format

The deliverable is a single markdown file saved to /mnt/user-data/outputs/ and presented via present_files so it renders as an artifact in the sidebar. Markdown is used at the document level for readability (H2 section headers, dividers, character-count callouts in italics). The prose content inside each SciENcv field is plain text only, with no markdown formatting (no bold, italics, bullets, or headers inside the field text). The user copies the prose from the rendered sidebar artifact directly into SciENcv.

Why this split: SciENcv strips most inline formatting and may reject inline markdown characters that survive copy-paste (asterisks, underscores), but it accepts paragraph breaks in the Personal Statement and Contributions to Science fields. A markdown-wrapped document gives the user a readable sidebar view while keeping the pastable content clean.

The file should be named nih_biosketch_<short_grant_descriptor>_draft.md where <short_grant_descriptor> is a one-or-two-word handle the user will recognize (e.g., the project name, PI last name, or mechanism). Examples: nih_biosketch_netGraph_draft.md, nih_biosketch_CEGS_draft.md, nih_biosketch_Farber_R01_draft.md.

Use this structure:

# NIH Biographical Sketch (SciENcv) draft

Field content for each SciENcv section is below. The prose inside each field is plain text and can be copy-pasted directly into SciENcv. Section headers and character-count callouts are document-level markdown for readability only and should not be pasted into SciENcv fields.

---

## C.1 Products closely related to the proposed project (up to 5)

1. [citation in CV format]
2. ...

---

## C.2 Other significant products (up to 5)

1. ...

---

## Supplement A: Personal Statement

*Characters: 3,247 / 3,500*

[Plain-text paragraphs. Paragraph breaks are allowed in SciENcv's Personal Statement field.]

---

## Supplement B: Honors (up to 15)

1. Honor: ... | Organization: ... | Year: ... | End Year: ...

---

## Supplement C: Contributions to Science (up to 5)

### Contribution 1: [Short descriptive title]

*Characters: 1,872 / 2,000*

*Inferred specific role (confirm or correct): [role inference summary]*

[Plain-text paragraph covering historical background, central findings, influence, and the user's specific role. Paragraph breaks allowed if needed.]

### Contribution 2: ...

Character counts are computed with len(text) where text is the prose content of the field (excluding the header line, the *Characters: ...* line, and the *Inferred specific role ...* line). Use code execution to count; do not estimate.

If a field exceeds its character limit, flag it loudly in the count line (e.g., *Characters: 2,134 / 2,000 — OVER LIMIT by 134*) and propose specific cuts at the bottom of that field. Do not silently exceed the limit.

After writing the markdown file, call present_files so it renders in the sidebar.

Section-by-section instructions

C.1 Products Closely Related to the Proposed Project

Select up to 5 products from the CV that are most relevant to the proposed project. Relevance criteria in priority order:

1. Direct topical overlap with the disease, system, or scientific domain in the grant
2. Use of techniques, methods, or computational approaches called for in the grant
3. Recent (last 5 to 10 years) preferred, unless an older product is uniquely relevant

Acceptable product types per NIH: publications, conference papers, presentations, websites, technologies, techniques, inventions, patents, patent applications, licenses, datasets, databases, software, models, educational aids, curricula, instruments, equipment, research materials, interventions, and new business creation.

Output each product using the citation format already present in the CV. Do not reformat or rewrite citations. Number them 1 through 5.

Three strong products beat five mediocre ones. If only 3 products in the CV have clear relevance to the grant, list 3 and note that explicitly.

C.2 Other Significant Products

Select up to 5 additional products from the CV that highlight Contributions to Science but were not picked for C.1. These do not need to relate to the proposed project.

Selection criteria:

• High-impact publications (high citation count if stated in the CV, high-impact venue, or seminal in their subfield)
• Products that anchor each of the user's distinct contribution areas (these usually map onto entries in Supplement C)
• Other products that demonstrate scientific influence: widely-used software, key datasets, major patents

Do not duplicate any product already listed in C.1.

Supplement A: Personal Statement

Limit: 3,500 characters including spaces. Plain text. Paragraph breaks are allowed (SciENcv accepts them in this field) but no inline markdown, bullets, or other formatting.

Briefly describe why the user is well-suited for their role on this project. Draw from a mix of:

• Relevant training (degrees, postdoctoral experience, specialized training)
• Prior or ongoing work on this specific topic or directly related topics
• Technical expertise that maps onto the grant's approach
• Collaborators or the scientific environment where the work will happen
• Past performance, including 1 to 3 ongoing or recently completed research projects from the past three years, with brief outcomes

In-text parenthetical references to any of the 10 products in C.1 plus C.2 are allowed. Use PMID, PMCID, or DOI if listed in the CV. Otherwise use (Author et al., Year). Do not include full bibliographic citations. Do not include hyperlinks.

The Personal Statement is argumentative, not summary. It makes the case that this person is the right fit, with specifics.

Tone: first-person, professional, concrete. The user has explicit writing preferences. Honor them:

• No em dashes. Use commas or parentheses.
• No filler ("Here's the thing", "It's worth noting", "In conclusion").
• No AI vocabulary: delve, tapestry, landscape, nuanced, utilize, leverage, streamline, ecosystem, paradigm, navigate, unpack, certainly, fundamentally, quietly.
• No "Not X. Y." contrast structures. No rhetorical questions answered in the next sentence. No stacked tricolons.
• No "serves as," "stands as," "represents." Use "is."
• Vary sentence length. Mix short and long. Avoid stacked short fragments.
• Be specific: name the grant, the disease, the technique, the collaborator, the dataset.

Supplement B: Honors

Pull honors directly from the CV. Each entry has three fields: Honor, Organization, Year.

Ranking when more than 15 honors exist in the CV:

1. Relevance to the project's topic or methodology
2. Recency (tiebreaker)
3. Prestige (further tiebreaker)

Cap at 15. SciENcv enforces this limit.

If the CV has few or no relevant honors, list what is in the CV without padding. Sometimes there are not many relevant honors to report, and that is fine.

If the CV has no honors at all, output:

1. Honor: Nothing to Report | Organization: Nothing to Report | Year: [current year]

After producing the list, ask the user to confirm or edit. They may want to swap in something not in the CV that they want to add. If they do, prompt them to also add it to their CV for consistency.

Supplement C: Contributions to Science

Up to 5 contributions. Each is limited to 2,000 characters. Plain text. Paragraph breaks allowed if they aid clarity, but typically each contribution is a single paragraph. No inline markdown, bullets, or headers inside the field text.

Each contribution must include all four components:

1. Historical background that frames the scientific problem
2. The central finding(s)
3. The influence of the finding(s) on the progress of science or application to health or technology
4. The user's specific role in the described work

For component 4 (specific role), infer from the CV first:

• Last or corresponding authorship usually signals senior author or PI role
• First authorship usually signals the user led the work
• Co-first authorship signals shared lead
• Middle authorship usually signals a contributor role (specific contribution if mentioned in the CV)
• For software, datasets, or patents: look for "Lead developer," "PI," "Co-inventor," etc. in the CV

After drafting all five contributions, list the inferred role for each one and ask the user to confirm or correct before treating the draft as final.

Organize contributions to span the user's distinct lines of work, not to all reinforce one project. Common patterns to draw from:

• Methodological contributions (new techniques, software, frameworks)
• Discovery contributions (specific scientific findings)
• Applied or translational contributions (clinical, policy, industry)
• Training, infrastructure, or community contributions, if scientifically substantive

In-text references to up to 5 of the 10 products in the Products section are allowed per contribution. Same citation format rules as the Personal Statement (PMID/PMCID/DOI if in the CV, otherwise Author et al., Year).

Manuscripts under development can be mentioned only if they appear in the CV with an explicit in-prep, in-review, or under-revision designation.

No figures, tables, or graphics (SciENcv does not allow them).

The same tone and language rules from the Personal Statement apply here.

Workflow

1. Check uploaded files. If CV or grant documents are missing, ask and stop. If multiple CVs are uploaded, ask which one to use and stop.
2. Read the CV in full. Read all grant documents in full.
3. Identify the proposed project's core: the disease or system, the methods, the central hypothesis or aims.
4. Draft C.1 and C.2 first. These anchor the products that the supplements will reference.
5. Draft Supplement A (Personal Statement).
6. Draft Supplement B (Honors).
7. Draft Supplement C (Contributions to Science). Note inferred specific role for each.
8. Run character counts (use code execution; do not eyeball).
9. Write the full markdown document to /mnt/user-data/outputs/nih_biosketch_<grant_descriptor>_draft.md with section dividers and character counts, then call present_files to surface it in the sidebar.
10. After delivering the draft, ask the user to:
    • Confirm or correct each inferred role in Contributions to Science
    • Confirm or edit the Honors list
    • Flag any section they want rewritten, reframed, or recompressed

Common failure modes to avoid

• Padding sections with vague claims to fill the character budget. Specificity beats density.
• Picking C.1 products that have only loose thematic ties to the grant. List 3 strong ones rather than 5 weak ones.
• Repeating the same products across multiple Contributions to Science. Each contribution should anchor on different products where possible.
• Writing a Personal Statement that reads like a CV summary. It should make a case.
• Inventing impact claims, awards, citation counts, or media coverage not in the CV.
• AI-tell language in the prose sections. Stephen has explicit deslop preferences in his user preferences; apply them.
• Missing one of the four required components in a Contribution to Science (historical background, central finding, influence, specific role). All four must be there.
• Producing markdown formatting inside the long-text fields. SciENcv strips it and the result looks broken.