atlassian-identity-cachelisted

# Atlassian identity cache Two Atlassian facts get looked up over and over by anything that calls the Atlassian MCP tools: the `cloudId` for a Jira site, and the `accountId` for a tracker user on that site. Both are constants for a given site, so rediscovering them on every call — or every iteration, for something like a headless loop — is pure waste. This skill resolves them through a small on-disk cache that heals itself if it's ever wrong, instead of a flat single value that would silently break the moment more than one Atlassian site is in play. ## Cache files - `$HOME/.claude/atlassian-cloud-ids.json` — `{ "<PROJECT-PREFIX>": "<cloudId>" }` - `$HOME/.claude/atlassian-account-ids.json` — `{ "<cloudId>": "<accountId>" }` Both are plain JSON maps. Treat a missing file as an empty map (`{}`), not an error — create it the first time you write to it. ## Resolving a cloud ID for a task key Given a task key like `PROJ-101`, the project-key prefix (`PROJ`) is everything before the `-`. 1. Look up the prefix in the cloud-ids map. 2. If it's there, use that `cloudId` directly for the tracker call you actually need to make — the call itself is the validation, so there's no separate check to pay for. - If the call succeeds, you're done. - If it fails as "issue not found" / "no permission" for that specific key, the cached value is stale: run **Resolve** below, overwrite the map entry with the result, then retry the original call once. 3. If the prefix isn't in the map at