Agent Skills › winstonkoh87/Athena-Public › academic-delivery

academic-delivery

GitHub

学术交付技能，通过8步流水线处理论文、报告等作业。核心在于强制自动触发红队审查，评估反方观点与形式分析，防止初稿直接交付，确保最终成果符合评分标准与学术规范。

examples/skills/workflow/academic-delivery/SKILL.md winstonkoh87/Athena-Public

Trigger Scenarios

需要撰写学术论文或分析报告涉及研究性写作或复杂问题解答

Install

npx skills add winstonkoh87/Athena-Public --skill academic-delivery -g -y

More Options

Non-standard path

npx skills add https://github.com/winstonkoh87/Athena-Public/tree/main/examples/skills/workflow/academic-delivery -g -y

Use without installing

npx skills use winstonkoh87/Athena-Public@academic-delivery

指定 Agent (Claude Code)

npx skills add winstonkoh87/Athena-Public --skill academic-delivery -a claude-code -g -y

安装 repo 全部 skill

npx skills add winstonkoh87/Athena-Public --all -g -y

预览 repo 内 skill

npx skills add winstonkoh87/Athena-Public --list

SKILL.md

Frontmatter

{
    "name": "academic-delivery",
    "cluster": "13 (Build Lifecycle) + 8 (Adversarial QA)",
    "created": 1772755200,
    "version": "1.0.0",
    "triggers": [
        "assignment",
        "essay",
        "report",
        "capstone",
        "academic",
        "TMA",
        "deliverable",
        "submission"
    ],
    "description": "8-step pipeline for academic deliverables — essays, reports, analysis, capstones. Encodes the full intake-to-delivery workflow with automatic red-team triggering.",
    "context_trigger": "*.docx, *.pdf, essay, report, assignment, capstone, coursework, academic, university, grade, APA, Harvard, MLA, literature review, methodology, research paper"
}

Academic Delivery Skill

Purpose: Structured pipeline for academic/knowledge deliverables. Prevents the V1-is-weak failure mode by auto-triggering adversarial review. Origin: Extracted from real-world execution of client assignments. Every assignment manually reconstructed this pipeline — now it's codified.

The 8-Step Pipeline

Step 1: INTAKE

Parse brief/requirements document
Extract: word count, format (essay/report/problem set), rubric criteria, deadline, client name
Identify prescribed frameworks or sources (e.g., "use Hirshfield as critical lens")
Log to .context/client_work/pricing_log.md if commercial

Step 2: SCOPE

Classify deliverable type:
- Essay (argumentative, comparative, reflective)
- Report (technical, research, capstone)
- Problem Set (calculations, code, SPSS)
- Presentation (slides, pitch deck)
Estimate complexity (Λ score)
If commercial: trigger client-pricing skill for quote generation

Step 3: RESEARCH

Load domain context via Exocortex (smart_search.py)
If external research needed: trigger deep-research-loop (Cluster #12)
Extract key frameworks, models, citations
Build a reference spine (3–7 sources minimum for essays)

Step 4: DRAFT (V1)

Write full first draft to spec
Embed citations inline
Target 90–95% of word count (leave room for red-team additions)
DO NOT deliver V1. V1 is always a working draft, never the output.

Step 5: RED-TEAM (Mandatory — Auto-Triggered)

[!IMPORTANT] This step is non-negotiable. It fires automatically after Step 4. Shipping V1 with zero counter-readings and zero formal analysis is the canonical failure mode.

Feed V1 through adversarial review (Cluster #8 — red-team-review)
Evaluate against:
1. Counter-reading: Does the draft contain at least one steelmanned opposing interpretation + rebuttal? (~80 words, non-negotiable)
2. Formal analysis: For literary/theoretical work — does it engage with the form (enjambment, structure, methodology), not just content?
3. Evidence density: Every claim backed by textual evidence or citation?
4. Rubric alignment: Does the draft hit every criterion in the brief?
5. Genre compliance: Is the output in the correct academic register? (MLA vs APA vs report format)
Generate a fix list with accept/reject decisions for each criticism
Reject invalid criticisms (category errors, phantom rubric scoring, n=1 methodology applied to literary analysis)

Step 6: REVISE (V2+)

Incorporate accepted fixes from red-team
If >3 structural issues found, produce V3 (rare — V2 is usually sufficient)
Re-check word count against target
Verify counter-reading and formal analysis are present

Step 7: COMPILE

Format per submission requirements:
- Cover page (if required)
- Table of Contents (forces a structural audit — per Session 06 learning)
- Section headers (topic-based, not numbered, for MLA essays)
- Works Cited / References (MLA, APA, or Harvard as specified)
- Appendices (if applicable)
Compaction triage: If over word count, cut from infrastructure (setup/transition/repetition) first, analytical core last. Ratio: 70% infrastructure cuts, 30% analysis cuts.
Paragraphing pass — wall-of-text paragraphs are a delivery failure, not a content failure.

Step 8: DELIVER

Final proofread (grammar, citation format, page numbers)
Export to required format (Markdown → Google Doc → DOCX)
If commercial: send to client with scope confirmation
Log completion to activeContext / quicksave

Exit Gate

No deliverable leaves Step 8 without:

Counter-reading present (if argumentative/analytical)
Formal analysis present (if literary/theoretical)
Word count within ±5% of target
All rubric criteria addressed
Format compliant (citations, headers, cover page)

Reflexion Archive

[REFLEXION] What failed: V1 shipped with zero counter-readings and zero formal analysis. Why: Over-optimised for clean thesis confirmation. Lesson: Auto-trigger red-team after V1. 80 words for a counter-reading is non-negotiable.

[REFLEXION] What failed: A capstone delivered at $83/hr effective rate (73% rate collapse vs benchmark). Why: No pricing skill fired during intake. Lesson: Step 2 (SCOPE) must trigger pricing evaluation for commercial work.