atlaslisted
Install: claude install-skill Udhay-Adithya/skills
# Atlas
Work too large for one session lives on the issue tracker, not in a file in the repository
and not in a conversation. The tracker is the shared state: any agent, in any harness, on
any machine, reads it and knows what is done, what is ready, and what is blocked.
This produces **decisions, not deliverables**, by default. Each issue resolves a question.
The map is finished when nothing is left to decide and someone can go and build. The urge
to just build the thing is usually the signal that the map is complete.
## Refer to issues by name
Every issue has a title. In anything a person reads, use the title and wrap the link
around it. Never a bare number.
A wall of `#42, #43, #44` is unreadable and a paragraph of titles is not. The numbers do
not disappear, they ride inside the link.
## The tracking issue
One issue, labelled `atlas:map`, is the canonical artifact. It is an **index, not a
store**: each decision lives in exactly one place, its own issue, and the map gists it and
links to it.
```markdown
## Destination
<what reaching the end of this looks like: the spec, the decision, the change. One or
two lines. Every session reads this before picking anything up.>
## Notes
<the domain, standing preferences for this effort, which skills each session should
invoke>
## Decisions so far
- [<issue title>](<url>): <one-line gist of the answer>
## Not yet specified
<questions that are coming but cannot be phrased sharply yet>
## Out of scope
<work consciously ru