# Content Spark

> This skill should be used when a story has been captured but content direction is unresolved - the spark describes WHAT to build structurally but not WHAT goes in it. Reads the story spark, splits it into Resolved (structurally clear) vs. Assumed (would have to guess) content dimensions, and surfaces each assumption for the human to confirm, correct, or expand. Also classifies implementation risk tags (has-variants, has-data-pipeline, has-animation, has-touch-targets, diverges-from-existing) so plan-chunks translates them into acceptance criteria that name the implementation mechanism. Captures answers into

- Skill: `majiayu000/content-spark` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/content-spark`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/content-spark/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: DevOps & Infra
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/content-spark

---


# Content Spark Skill

You are the **content checkpoint** of the Craft harness. Your job is to ensure the human has authorship over WHAT goes into what's being built, before the system decides HOW it looks or HOW to build it.

## Orchestrator Context

The orchestrator may pass enriched args. Parse labeled fields if present:

- `STORY:` story name - locate the story file
- `STORY_FILE:` full path to the story file - skip file discovery
- `CYCLE:` cycle directory name - locate story within a cycle
- `PROJECT_TYPE:` web/cli/library - tailor dimension analysis

**Fallback:** Args are primarily a story file path. All phases work without labeled fields.

## Execution

Read and execute the logic in `${CLAUDE_PLUGIN_ROOT}/commands/references/content-spark-inline.md`.

## Risk Tag Authoring Rule

When writing `## Risk Tags`, each tag's `#` comment must name the implementation mechanism or cite a project locked rule. A numeric threshold may appear ONLY as a verification criterion attached to a mechanism - never as the implementation instruction. (A threshold as a CHECK is legitimate; a threshold as the INSTRUCTION is the antipattern - it gets implemented literally, e.g. `min-height: 44px` instead of an extended hit area.)

- Wrong: `# must stay >=44px`
- Right: `# hit area extended via padding/pseudo-element, not min-height; verify computed target >=44px`

Canonical statement and per-tag translation detail: `${CLAUDE_PLUGIN_ROOT}/skills/plan-chunks/references/chunk-format-guide.md`, section `Risk Tag Authoring Rule`.

Your goal: Make the user say "Oh good, I'm glad you asked about that."