Agent Skills: Write Skill

Write blog posts from session insights with multi-perspective analysis.

UncategorizedID: jongwony/epistemic-protocols/write

Install this agent skill to your local

pnpm dlx add-skill https://github.com/jongwony/epistemic-protocols/tree/HEAD/epistemic-cooperative/skills/write

Skill Files

Browse the full folder contents for write.

Download Skill

Loading file tree…

epistemic-cooperative/skills/write/SKILL.md

Skill Metadata

Name
write
Description
Write blog posts from session insights with material recall and multi-perspective analysis.

Write Skill

Transform session insights and conceptual explorations into structured, publishable content through multi-perspective analysis and iterative refinement.

Invoke directly with /write when the user wants blog drafting from session insights.

When to Use

Invoke this skill when:

  • Converting framework development or analytical sessions into blog posts
  • Writing about conceptual topics requiring multi-perspective review
  • Structured iterative refinement anticipated

Skip when:

  • Technical tutorials with code (standard writing)
  • Documentation updates (direct Edit)
  • Single-pass content without review need

Workflow Overview

RECOLLECT → PROTHESIS(Context→Perspective→Inquiry→Synthesis) → FORMAT → DRAFT → REFINE → VALIDATE → FINALIZE

| Phase | Tool | Decision Point | |-------|------|----------------| | Recollect | /recollect protocol | Prior session material recognition | | Prothesis | /frame protocol | Context-derived perspectives, parallel inquiry | | Format | AskUserQuestion | Output type, language | | Draft | Write | — | | Refine | Edit (loop) | User feedback | | Validate | /gap | Gap detection | | Finalize | Edit | — |

Phase Execution

0. Material Recall (/recollect)

Invoke /recollect to surface prior session material relevant to the drafting topic — terminology coined in earlier sessions, prior decisions referenced, framings already established. The recalled context enriches the Phase 1-3 context acquisition (G(U) → C), improving perspective selection in the multi-perspective analysis.

This phase is conditional: skip silently when the drafting topic has no prior session footprint (e.g., entirely new domain). Anamnesis's internal Phase 0 scan determines applicability.

1-3. Prothesis Protocol (Multi-Perspective Analysis)

Apply the Prothesis protocol (/frame) for epistemic analysis:

Phase 0: G(U) → C              -- Context acquisition from session/topic
Phase 1: C → {P₁...Pₙ}(C) → Pₛ -- Perspectives derived FROM context (not predefined)
Phase 2: Pₛ → ∥I(Pₛ) → R       -- Parallel inquiry with Horizon Limits
Phase 3: R → Syn(R) → L        -- Synthesis: convergence, divergence, assessment

Key differences from standalone /frame:

  • Continues automatically to Format phase after Synthesis
  • Lens L becomes input for content generation

Reference: prothesis/skills/frame/SKILL.md

4. Format Decision

Present output options:

  • Blog Post (Korean/English)
  • Essay
  • Newsletter
  • Thread

5. Draft Generation

Write initial draft to ~/.claude/.write/ directory:

  • Filename: YYYY-MM-DD-{topic-slug}.md
  • Structure: Hook → Context → Framework → Application → Implications

6. Iterative Refinement

Loop on user feedback:

  • Incremental changes → Edit directly
  • Structural changes → Generate option versions (A, B, C)

Exit conditions:

  • User approval
  • Explicit "finalize" command

7. Gap Detection

Invoke /gap for final validation:

  • Procedural gaps
  • Consideration gaps
  • Duplicate content

Composition note: When /write output is downstream-reviewed by /comment-review, this Phase 7 internal /gap and /comment-review's Scan Step 2 /gap address distinct scopes (internal draft quality here vs. fixation-event decision quality downstream). Both invocations are intentional; the scope distinction preserves the /gap invariant.

8. Finalization

Apply final edits. Optionally clean intermediate versions.

Quality Criteria

| Metric | Limit | |--------|-------| | Concepts per section | ≤3 | | Framework components | ≤5 | | Abstraction layers | ≤2 |

Content Transformation

What Becomes Content

| Session Element | Blog Element | |-----------------|--------------| | Problem context | Opening hook | | Multi-perspective analysis | Framework structure | | Convergence points | Core thesis | | Divergence points | Discussion sections | | Resolution approach | Actionable methodology |

What Gets Filtered

  • Tool invocations, command outputs
  • Trial-and-error debugging steps
  • Redundant restatements
  • Context-specific details (paths, configs)

Integration

This skill integrates with:

  • /recollect — Material recall (Phase 0)
  • /frame — Multi-perspective analysis (Phases 1-3)
  • /gap — Gap detection (Phase 7)

Additional Resources

For detailed workflow steps and content transformation rules:

  • references/workflow.md — Complete phase descriptions