feat(queue): durable local close queue with rate-limit retry (RUSH-2308)

CHANGELOG.md

@@ -5,6 +5,36 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [Unreleased]
+
+### Added
+
+- **`linear queue` / `linear queue drain`** — durable local queue for closes
+  that cannot be written immediately. `linear update <id> --done --proof ...`
+  now persists the intent before attempting the API; on a Linear rate limit
+  (429) or transient error the intent is retained in `~/.linear-cli/queue/`
+  and retried later. `linear queue drain` applies pending closes with
+  exponential backoff; `linear queue drain --dry-run` previews them. The next
+  `linear update --done` automatically drains the queue first.
+- **Idempotent queued closes.** Duplicate intents for the same ticket collapse
+  to the latest proof/comment. A drain that finds the issue already in the
+  desired state removes the intent without re-posting proof.
+- **Bounded queue growth.** `MAX_QUEUE_SIZE` caps the number of distinct
+  intents; new intents are rejected when full, but existing intents can still
+  be updated. `MAX_QUEUE_ATTEMPTS` limits retries so stuck items do not linger
+  forever.
+
+### Changed
+
+- `gql` now surfaces HTTP status codes in GraphQL error extensions so callers
+  can distinguish rate limits (429) and transient errors from permanent
+  failures.
+
+### Docs
+
+- README and `skill.md` document `linear queue`, `linear queue drain`, and the
+  durable-close behavior.
+
 ## [0.17.0] - 2026-08-06
 
 ### Added

README.md

@@ -93,6 +93,10 @@ linear update ANT-1 ANT-2 ANT-3 --cycle none             # bulk: many ids at onc
 linear tasks --cycle all --json | jq -r '.issues[].identifier' \
   | linear update --stdin --label triage                 # bulk via stdin (xargs-style)
 
+linear queue                            # list closes waiting on rate limits
+linear queue drain                      # apply queued closes with backoff
+linear queue drain --dry-run            # preview queued closes without applying
+
 linear create "Fix auth bug" --label security --priority high
 linear create --description "Paragraph dump — title is derived from this."
 linear create "Sub-task" --parent ANT-42     # nested; prints a tip nudging a flat issue
@@ -143,6 +147,7 @@ The same CLI works whether you're typing or a subagent is. Driving Linear from e
 - **Native agent delegation.** `linear update ANT-42 --delegate claude` sets Linear's `delegateId`: the human stays assignee, the agent becomes delegate, and review ownership stays clear.
 - **One ownership model.** `delegate` is the only thing that owns an issue. `linear tasks --agent claude` filters to issues delegated to Claude; the default view adds the issues nobody has been delegated (`delegate` is null). `linear tasks --board` groups its columns by delegate. There is no label lane — an unknown `--agent` aborts rather than printing an empty queue.
 - **Proof-first completion.** `--done --proof <file|url|text>` uploads attachments, records links, and appends notes in one call — so reviewers see evidence without digging.
+- **Durable closes.** If a `--done` call hits a Linear rate limit or transient error, the close intent is persisted to `~/.linear-cli/queue/` and retried with exponential backoff. `linear queue drain` applies pending closes; the next `linear update --done` also drains automatically so the board stays honest.
 - **JSON everywhere.** `--json` on every read command. Pipe to `jq` or hand to a subagent.
 
 <p align="center">