Academic Writing

discussion-composer

Composes Discussion sections around key findings, mechanisms, clinical relevance, and limitations. Polished: citation placeholder density rule (max 4 per 400 words) added; discussion length calibration by manuscript type added; dead references removed.

83100Total Score

Core Capability

86 / 100

Functional Suitability

11 / 12

Reliability

11 / 12

Performance & Context

7 / 8

Agent Usability

15 / 16

Human Usability

7 / 8

Security

12 / 12

Maintainability

6 / 12

Agent-Specific

17 / 20

Medical Task

14 / 15 Passed

85Retrospective cohort study: HR 1.43 (95% CI 1.12-1.82), one prior study provided for comparison

5/5

82Key findings provided but no prior literature — skill must use citation placeholders throughout

4/5

76Minimal input: 'Write my discussion' — no findings, no research question

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

Dimension	Result	Detail
Scientific Integrity	PASS	No fabricated citations, studies, or findings detected. Hard rules mandate citation placeholders when prior literature is not provided.
Practice Boundaries	PASS	No clinical recommendations beyond what evidence explicitly supports. Skill scope is discussion prose writing only.
Methodological Ground	PASS	No methodological fallacies. Limitations rule (constraint → impact → mitigation) enforces honest and proportionate acknowledgment.
Code Usability	N/A	No code generated; Mode A text-output skill.

Core Capability86 / 100 — 8 Categories

Functional Suitability

6-part structure comprehensive; citation placeholder rule and limitations constraint→impact→mitigation format are practically valuable; no handling for multi-study manuscripts or meta-analysis discussion structure.

11 / 12

92%

Reliability

Step 3 revision checklist with 8 verifiable criteria is strong; citation placeholder explicitly prevents fabrication; ask-before-writing rule well-enforced.

11 / 12

92%

Performance & Context

Clean 4-step workflow with 3-part output (draft + placeholder note + assumptions); minor verbosity in optional inputs list.

7 / 8

88%

Agent Usability

Good sample triggers, 6-part template, and citation placeholder mechanism; feedback design via 3-part output and Step 3 checklist effective.

15 / 16

94%

Human Usability

Rich trigger phrase list; scope boundary (not Introduction, Methods, or Results) clearly stated.

7 / 8

88%

Security

Full marks. Hard rules prohibit fabricated citations, invented prior studies, and clinical recommendations beyond evidence.

12 / 12

100%

Maintainability

Only audit-reference.md present (an audit config, not a modular rule file); dead references to references/guide.md and references/examples/ directory; all rules embedded inline in SKILL.md with no modular structure.

6 / 12

50%

Agent-Specific

Trigger precision and citation placeholder composability strong; dead reference assets reduce maintainability and confidence in completeness.

17 / 20

85%

Core Capability Total86 / 100

Medical TaskExecution Average: 81 / 100 — Assertions: 14/15 Passed

Canonical

Retrospective cohort study: HR 1.43 (95% CI 1.12-1.82), one prior study provided for comparison

5/5 ✓

Variant A

Key findings provided but no prior literature — skill must use citation placeholders throughout

4/5 ✓

Edge

Minimal input: 'Write my discussion' — no findings, no research question

5/5 ✓

Canonical✅ Pass

Retrospective cohort study: HR 1.43 (95% CI 1.12-1.82), one prior study provided for comparison

5/5 assertions passed. Complete 6-part discussion with correct citation, hedged language, and limitations in required format.

Basic 35/40|Specialized 50/60|Total 85/100

✅A1Output follows 6-part discussion structure (opening, interpretation, literature comparison, implications, limitations, conclusion)

✅A2Provided prior study is cited and compared, not ignored

✅A3Limitations are formatted as constraint → impact on interpretation → mitigation/future direction

✅A4Hedged language ('suggests', 'may reflect', 'is consistent with') used throughout interpretation

✅A5No new data or results introduced in the Discussion that were not in the user's provided results

Pass rate: 5 / 5

Variant A✅ Pass

Key findings provided but no prior literature — skill must use citation placeholders throughout

4/5 assertions passed. Citation placeholders used correctly; placeholder density slightly high for the discussion length.

Basic 33/40|Specialized 49/60|Total 82/100

✅A1Output uses [CITE: ...] placeholders rather than invented citations

✅A2Citation placeholders are specific enough to guide the user's literature search

✅A3Step 4 output explicitly notes all placeholders inserted and how many need to be filled

❌A4Placeholder density is proportionate to discussion length (not excessive relative to prose)

✅A5Discussion remains readable and interpretable despite placeholders

Pass rate: 4 / 5

Edge✅ Pass

Minimal input: 'Write my discussion' — no findings, no research question

5/5 assertions passed. Input collection triggered correctly; no discussion drafted.

Basic 30/40|Specialized 46/60|Total 76/100

✅A1Skill does not draft discussion without key results and research question

✅A2Output asks specifically for primary finding (with quantitative detail) and research question

✅A3Output also asks about tone/depth preference (brief vs full discussion)

✅A4Output does not invent findings or fabricate a placeholder discussion

✅A5Output communicates the 6-part structure that will be applied once inputs are provided

Pass rate: 5 / 5

Medical Task Total81 / 100

Key Strengths

Citation placeholder rule ([CITE: study showing...]) prevents fabrication while keeping the discussion structurally complete — a practical and safe design choice for a writing skill
Limitations format (constraint → impact → mitigation) provides concrete structural guidance that prevents the common failure of listing limitations without explanation
Step 3 revision checklist with 8 verifiable criteria enables systematic self-audit before delivery
Clear distinction between Discussion scope and other section types (Introduction, Methods, Results) prevents scope creep