story-startlisted
Install: claude install-skill vitdo82/story-drift
Argument: a Jira key, e.g. `/story-start PROJ-237`
## 1. Read
Via the Atlassian MCP: the story, its parent epic, every linked story, **every sibling story under that epic**, and all comments. If the epic description contradicts the story, say so — the epic wins until the PO says otherwise.
## 2. Write the snapshot
`.story/<KEY>/context.json` — this is what `/story-sync` compares against later, so the normalisation must be stable: strip whitespace and markdown before hashing.
```json
{
"key": "PROJ-237",
"captured_at": "<iso8601>",
"fields": {
"summary": {"value": "...", "hash": "<sha256>"},
"description": {"value": "...", "hash": "<sha256>"},
"acceptance_criteria": {"value": "...", "hash": "<sha256>"}
},
"epic": {"key": "PROJ-100", "description_hash": "...", "priority": "..."},
"linked": [{"key": "PROJ-201", "status": "..."}],
"comments": [{"id": "...", "author": "...", "created": "...", "body_hash": "..."}],
"epic_tree": [
{"key": "PROJ-201", "status": "...", "summary_hash": "...",
"description_hash": "...", "ac_hash": "...", "last_comment_id": "..."}
]
}
```
`epic_tree` must list **every child of the parent epic**, not just the linked ones. A sibling changing its acceptance criteria can change your contract.
## 3. Write the tech design
Markdown, into the repo's docs directory (not an HTML attachment on the ticket — those diff badly and agents read them worse).
Sections: Context · **Requirements baseline** (one row per acceptan