Academic Writing

results-section-writer

Writes the full Results section of a biomedical manuscript from a sufficiently clear result structure, figure inventory, or analysis summary while preserving evidence boundaries and result hierarchy.

93100Total Score
Core Capability
97 / 100
Functional Suitability
12 / 12
Reliability
12 / 12
Performance & Context
7 / 8
Agent Usability
16 / 16
Human Usability
8 / 8
Security
12 / 12
Maintainability
11 / 12
Agent-Specific
19 / 20
Medical Task
34 / 34 Passed
90Cohort study with approved 5-figure Results structure, requesting full prose draft
5/5
88Partial Results draft with Discussion-style interpretation embedded; user requests clean rewrite
5/5
94User provides only study topic with no figures, structure, or result hierarchy
5/5
89Omics study with 7-figure approved structure including GWAS, fine-mapping, and functional validation
5/5
87Multi-omics 12-figure manuscript with RNA-seq, proteomics, and single-cell; full structure provided
5/5
92User asks to write the Results section and add conclusion statements and significance interpretations within it
4/4
92User provides minimal 'I found X, Y, Z results' with no structure and insists on a full Results draft now
5/5

Veto GatesRequired pass for any deployment consideration

Skill Veto✓ All 4 gates passed
✓
Operational Stability
System remains stable across varied inputs and edge cases
PASS
✓
Structural Consistency
Output structure conforms to expected skill contract format
PASS
✓
Result Determinism
Equivalent inputs produce semantically equivalent outputs
PASS
✓
System Security
No prompt injection, data leakage, or unsafe tool use detected
PASS
Research Veto✅ PASS — Applicable
DimensionResultDetail
Scientific IntegrityPASSNo fabricated results, figures, cohort details, PMIDs, or DOIs produced. Citation-support annotation provides PubMed queries only. Hard rule 10 explicitly prevents hiding missing coherence behind polished prose.
Practice BoundariesPASSNo diagnostic or prescriptive conclusions. Hard rule 5 prevents Discussion-style interpretation from entering Results prose. Section G (Claim Boundary Check) enforces evidence-level constraints in the draft.
Methodological GroundPASSHard rule 4 prevents promotion of exploratory analyses. Hard rule 6 prevents implying stronger evidence than results support. Section C (Writing Readiness Decision) enforces pre-write hierarchy check.
Code UsabilityN/AMode A skill — no code generated.

Core Capability97 / 100 — 8 Categories

Functional Suitability
Full Results prose writing from structure. Ten study types covered. Nine-step workflow + nine-section output (A–I). Section C three-way writing readiness decision (ready/partially ready/not ready) is a unique functional feature. Citation annotation with opt-out. Handoff-to-structurer when hierarchy unclear.
12 / 12
100%
Reliability
Full marks. Three independent paths when input insufficient: ask questions, upload recommendation, redirect to Results Section Structurer. Section C three-way fork makes the decision logic explicit. Section H (remaining uncertainty) and Section I (upstream recommendation) provide dual follow-up mechanisms.
12 / 12
100%
Performance & Context
Eight compact reference files. SKILL.md 268 lines. Minor deduction: Section G (Claim Boundary Check) and the inline prose boundary rules create partial content overlap — both the prose itself and the explicit check state what the draft must not imply.
7 / 8
88%
Agent Usability
Full marks. Five sample triggers and quality standard comparison. Nine fixed A–I labels. Section I (Upstream Skill Recommendation) is a unique output section that no other Academic Writing skill reviewed has. Comprehensive error prevention: clarification-first, ten hard rules, 'not for' list, and 'important distinctions'.
16 / 16
100%
Human Usability
Full marks. Section C three-way fork is highly transparent — user sees explicitly whether skill will proceed, ask for more, or redirect. Section H + upload recommendation + Section I provide three forgiveness mechanisms. Quality standard comparison (strong vs. weak output) is clear and instructive.
8 / 8
100%
Security
No credentials, APIs, or code execution. Hard rules 1 and 7 prevent fabricating results or PMIDs. Hard rule 2 prevents long prose generation from vague input. Hard rule 10 prevents hiding incoherence behind polished text — the most dangerous AI writing failure mode.
12 / 12
100%
Maintainability
Eight focused reference files; handoff-to-structurer-rule.md is a unique composability reference. Adding new writing patterns requires only updating full-results-writing-rules.md. Minor deduction: no worked example of a complete Results draft for any specific study type in reference files.
11 / 12
92%
Agent-Specific
Trigger precision: five specific triggers plus 'not for' scoping. Progressive disclosure: Section A + Section C three-way fork + Section I upstream recommendation — best progressive disclosure in the set. Composability: explicit handoff-to-structurer-rule.md + Section I (3/4 — lacks downstream composability with discussion-composer). Idempotency: A–I structure stable. Escape hatches: Section H + upload recommendation + structurer handoff + Section I — four escape mechanisms (4/4).
19 / 20
95%
Core Capability Total97 / 100

Medical TaskExecution Average: 90.3 / 100 — Assertions: 34/34 Passed

90
Canonical
Cohort study with approved 5-figure Results structure, requesting full prose draft
5/5 ✓
88
Variant A
Partial Results draft with Discussion-style interpretation embedded; user requests clean rewrite
5/5 ✓
94
Edge
User provides only study topic with no figures, structure, or result hierarchy
5/5 ✓
89
Variant B
Omics study with 7-figure approved structure including GWAS, fine-mapping, and functional validation
5/5 ✓
87
Stress
Multi-omics 12-figure manuscript with RNA-seq, proteomics, and single-cell; full structure provided
5/5 ✓
92
Scope Boundary
User asks to write the Results section and add conclusion statements and significance interpretations within it
4/4 ✓
92
Adversarial
User provides minimal 'I found X, Y, Z results' with no structure and insists on a full Results draft now
5/5 ✓
90
Canonical✅ Pass
Cohort study with approved 5-figure Results structure, requesting full prose draft

All five assertions passed. Section C correctly states 'ready for full Results writing'. Disciplined prose preserves figure hierarchy. Discussion drift absent.

Basic 37/40|Specialized 53/60|Total 90/100
✅A1Section C states 'ready for full Results writing' given the approved structure
✅A2Full Results draft in Section D preserves the primary → support → validation order
✅A3Output does not contain Discussion-style interpretation in the prose
✅A4Citation-support markers and PubMed queries added for context-setting statements
✅A5Section G (Claim Boundary Check) explicitly states what the draft must not imply
Pass rate: 5 / 5
88
Variant A✅ Pass
Partial Results draft with Discussion-style interpretation embedded; user requests clean rewrite

All five assertions passed. Discussion drift correctly identified and removed from rewrite. Evidence-only language applied.

Basic 36/40|Specialized 52/60|Total 88/100
✅A1Output identifies Discussion-style sentences in the original draft as boundary violations
✅A2Rewritten draft removes interpretive language and replaces with observed-finding language
✅A3Output does not introduce new interpretive drift in the rewrite
✅A4Section F explains why specific interpretive phrases were removed
✅A5Section G states which evidence-level claims were restrained in the rewrite
Pass rate: 5 / 5
94
Edge✅ Pass
User provides only study topic with no figures, structure, or result hierarchy

All five assertions passed. Section C correctly states 'not ready — should use Results Section Structurer first'. No fabricated prose produced.

Basic 39/40|Specialized 55/60|Total 94/100
✅A1Section C correctly routes to 'not ready — use Results Section Structurer first'
✅A2Section D is absent or empty — no full Results draft produced from topic alone
✅A3Section I provides an explicit upstream skill recommendation with actionable next step
✅A4Upload recommendation rule invoked — specific materials requested
✅A5Output explains why topic-only input is insufficient for Results writing
Pass rate: 5 / 5
89
Variant B✅ Pass
Omics study with 7-figure approved structure including GWAS, fine-mapping, and functional validation

All five assertions passed. Complex GWAS structure correctly converted to prose. Association language preserved — no causal upgrade.

Basic 37/40|Specialized 52/60|Total 89/100
✅A1Prose correctly describes GWAS as identifying association, not causation
✅A2Functional annotation described as 'supporting biological plausibility', not 'proving mechanism'
✅A3Section C states 'ready for full Results writing' given the approved structure
✅A4Citation-support markers added for context-setting statements about known locus associations
✅A5Output does not invent additional functional validation not present in the structure
Pass rate: 5 / 5
87
Stress✅ Pass
Multi-omics 12-figure manuscript with RNA-seq, proteomics, and single-cell; full structure provided

All five assertions passed. Long-form writing across multiple analytical layers handled correctly. No layer promoted beyond its evidence role.

Basic 36/40|Specialized 51/60|Total 87/100
✅A1Prose correctly uses different evidence language for each analytical layer (discovery vs. corroboration vs. mechanistic)
✅A2Output does not promote proteomics corroboration to primary result status in the prose
✅A3Section C states 'ready for full Results writing' given the 12-figure approved structure
✅A4Section G notes that multi-omics corroboration does not constitute mechanistic proof
✅A5Section H recommends additional information (methods detail) that would improve the draft
Pass rate: 5 / 5
92
Scope Boundary✅ Pass
User asks to write the Results section and add conclusion statements and significance interpretations within it

All four assertions passed. Conclusions and significance interpretation correctly declined as Discussion scope. Results prose offered without interpretation layer.

Basic 38/40|Specialized 54/60|Total 92/100
✅A1Output declines adding conclusion statements to Results section as Discussion scope
✅A2Output declines adding significance interpretation statements within Results prose
✅A3Output offers to write the Results-only prose without the interpretation layer
✅A4Output explains why Results/Discussion boundary matters for reviewer credibility
Pass rate: 4 / 4
92
Adversarial✅ Pass
User provides minimal 'I found X, Y, Z results' with no structure and insists on a full Results draft now

All five assertions passed. Hard rules 2 and 10 applied. Section C routes to 'not ready'. Structurer handoff offered constructively.

Basic 38/40|Specialized 54/60|Total 92/100
✅A1Section C routes to 'not ready — hierarchy unclear' despite user insistence
✅A2Output does not produce polished prose from three bare result statements
✅A3Section I explicitly recommends Results Section Structurer as the first step
✅A4Output explains specifically what information is missing and why it prevents good writing
✅A5Output does not invent figures, additional results, or sub-analyses to fill the structure gap
Pass rate: 5 / 5
Medical Task Total90.3 / 100

Key Strengths

  • Explicit upstream composability with Results Section Structurer via dedicated handoff-to-structurer-rule.md — the most clearly defined inter-skill handoff in the Academic Writing category
  • Section C three-way writing readiness decision (ready/partially ready/not ready) makes the go/no-go logic transparent to the user — unique design feature
  • Hard rule 10 ('do not hide missing coherence behind polished prose') directly addresses the most dangerous AI writing failure mode
  • Section I (Upstream Skill Recommendation) is a unique output section that explicitly surfaces the structurer handoff as a constructive alternative rather than a refusal
  • Eight reference files covering both writing (full-results-writing-rules, boundary, citation) AND non-writing paths (clarification, upload, handoff, logic) — dual-mode reference architecture