Academic Writing
poster-storyline-builder
Reorganizes a paper into a storyline suitable for scientific posters. Plans section structure, title hierarchy, figure selection, and live-explanation flow for academic conference posters.
87100Total Score
Core Capability
87 / 100
Functional Suitability
11 / 12
Reliability
10 / 12
Performance & Context
8 / 8
Agent Usability
14 / 16
Human Usability
7 / 8
Security
12 / 12
Maintainability
8 / 12
Agent-Specific
17 / 20
Medical Task
14 / 15 Passed
90Phase 2 RCT poster for ASH: BTKi vs control in relapsed/refractory CLL (ORR 78% vs 45%, p<0.001, n=120, 36×48 inch portrait). Figures: KM curves, waterfall plot, AE table.
5/5
88Basic science poster for AACR: CRISPR screen identifying PARP1 as synthetic lethal partner with BRCA2 in TNBC (10-fold olaparib sensitization). Figures: volcano plot, cell viability curves, western blot.
4/5
84Sparse input: 'Help me make a poster about my research on immune cells and cancer'
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
PASSResearch Veto✅ PASS — Applicable
| Dimension | Result | Detail |
|---|---|---|
| Scientific Integrity | PASS | No fabricated results or conclusions. Hard rules explicitly prohibit recommending figures not mentioned in the user's study and fabricating results not in the source material. |
| Practice Boundaries | PASS | No clinical advice or prescriptive recommendations. Scope limited to poster layout planning. |
| Methodological Ground | PASS | No methodological claims. Poster layout advice is evidence-based communication design rather than scientific methodology. |
| Code Usability | N/A | No executable code generated; Mode A direct-execution skill. |
Core Capability87 / 100 — 8 Categories
Functional Suitability
5-step workflow covers story extraction, figure selection, text compression, layout plan, and design tips. Space allocation table and per-section word count targets are specific and actionable. Minor deduction: no mention of conference-specific poster format variations (portrait/landscape, A0/A1, digital vs physical) that meaningfully affect layout decisions.
11 / 12
92%
Reliability
Input validation correctly triggers follow-up questions when source material is too sparse. Hard rules prevent figure fabrication. Minor deductions: no explicit flagging when primary result is unquantified (affects figure selection quality), and font size recommendations are not verified against specific conference templates.
10 / 12
83%
Performance & Context
Full marks. SKILL.md is compact (~133 lines). Reference file is a stub (3 lines). Step 1 story extraction runs before layout plan generation, which is efficient.
8 / 8
100%
Agent Usability
Trigger phrases are specific and cover all major entry points. Poster section structure ASCII diagram is highly readable. Minor deductions: no confirmation step between story extraction and full layout plan delivery (user cannot correct Step 1 interpretation before Step 4 output), and no explicit guidance for when user provides too many figures to prioritize.
14 / 16
88%
Human Usability
Strong 'When to Use' section with 5 specific scenarios. Out-of-scope note about design tools prevents scope confusion. Minor deduction: no clear path described when user has only a partial manuscript (e.g., abstract only, no methods or figure list).
7 / 8
88%
Security
Full marks. Hard rules explicitly prohibit fabricating results and recommending unmentioned figures. No credential or injection risks.
12 / 12
100%
Maintainability
references/guidelines.md is a stub (3 lines, no substantive content). All functional content is inline in SKILL.md without modular separation. Low modularity — adding conference-specific guidelines would require editing SKILL.md directly. Significant maintainability deduction.
8 / 12
67%
Agent-Specific
Trigger precision is excellent. Progressive disclosure via 5 steps prevents output overload. Idempotent: same abstract → same layout plan. Minor deductions: no composability with graphical-abstract-generator or slide-deck-for-lab-meeting despite natural workflow adjacency; and escape hatch for unknown conference format is not explicitly described.
17 / 20
85%
Core Capability Total87 / 100
Medical TaskExecution Average: 87.3 / 100 — Assertions: 14/15 Passed
90
Canonical
Phase 2 RCT poster for ASH: BTKi vs control in relapsed/refractory CLL (ORR 78% vs 45%, p<0.001, n=120, 36×48 inch portrait). Figures: KM curves, waterfall plot, AE table.
5/5 ✓
88
Variant A
Basic science poster for AACR: CRISPR screen identifying PARP1 as synthetic lethal partner with BRCA2 in TNBC (10-fold olaparib sensitization). Figures: volcano plot, cell viability curves, western blot.
4/5 ✓
84
Edge
Sparse input: 'Help me make a poster about my research on immune cells and cancer'
5/5 ✓
90
Canonical✅ Pass
Phase 2 RCT poster for ASH: BTKi vs control in relapsed/refractory CLL (ORR 78% vs 45%, p<0.001, n=120, 36×48 inch portrait). Figures: KM curves, waterfall plot, AE table.
Step 1 correctly extracts all 5 story elements. Figure plan: KM curves (must-include primary result), waterfall plot (should-include), AE table (convert to visual bar chart or summarize in text). Layout plan with 4-column structure, word count targets per section, take-home sentence, and design tips produced.
Basic 37/40|Specialized 53/60|Total 90/100
✅A1Step 1 correctly identifies all 5 story elements (problem, objective, approach, answer, implication)
✅A2Figure plan recommends 2-4 figures maximum with clear must-include/should-include/cut guidance
✅A3Section plan includes word count targets per the text compression table
✅A4Take-home sentence is quantitative and specific (includes primary endpoint numbers)
✅A5No results or conclusions fabricated beyond what was provided in the input
Pass rate: 5 / 5
88
Variant A✅ Pass
Basic science poster for AACR: CRISPR screen identifying PARP1 as synthetic lethal partner with BRCA2 in TNBC (10-fold olaparib sensitization). Figures: volcano plot, cell viability curves, western blot.
Layout plan adapted for basic science: Methods column given more space than clinical RCT (novel CRISPR methodology). Volcano plot as primary results/methods figure, viability curves as main result, western blot as supporting. Minor: layout does not explicitly note that AACR-specific size requirements should be verified.
Basic 36/40|Specialized 52/60|Total 88/100
✅A1Story extracted correctly for basic science format (biological problem, mechanistic objective, experimental approach, key finding, translational implication)
✅A2Methods section allocated more poster space than standard clinical RCT layout due to novel CRISPR methodology
✅A3Figure plan prioritizes cell viability curves as primary result figure with volcano plot as context
❌A4Layout plan notes that AACR conference-specific poster format should be verified
✅A5No figures recommended beyond the three user-specified figures
Pass rate: 4 / 5
84
Edge✅ Pass
Sparse input: 'Help me make a poster about my research on immune cells and cancer'
Hard rule triggered: source material too sparse to create a layout. Focused questions asked covering primary result with numbers, available figures, conference/size, and primary audience. No layout plan produced from vague 'immune cells and cancer' topic.
Basic 35/40|Specialized 49/60|Total 84/100
✅A1Skill correctly withholds layout plan from topic-only sparse input
✅A2Focused questions cover primary result (with numbers), available figures, conference format, and audience type
✅A3No fabricated immune cell or cancer results generated from the vague input
✅A4Response explains what information is needed and why each piece matters for layout planning
✅A5Provisional section structure offered as a starting template for common oncology posters
Pass rate: 5 / 5
Medical Task Total87.3 / 100
Key Strengths
- Space allocation guidelines (Results 35-45%, Introduction 10-15%) are practically calibrated and rare in academic communication tools
- Figure selection strategy (Must-include/Should-include/Cut-for-poster) converts vague figure priority decisions into actionable guidance
- Text compression table with per-section word count targets gives users measurable goals rather than generic advice
- Design accessibility tip (colorblind-safe color scheme, avoid red/green) addresses a commonly overlooked poster quality issue