/ship on Codex starts its audits with Codex's own subagent tools, not Claude Code's (#4476) (#4489)

On Codex, /ship and the other skills now start audit and reviewer
subagents with Codex's own `spawn_agent` and wait on `wait_agent`,
instead of Claude Code's Agent tool, which Codex can't call.

Why: Codex's /ship stopped at the Step 7 coverage audit and never
finished (#4476).

Codex users need `multi_agent = true` under `[features]` in
`~/.codex/config.toml`.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
This commit is contained in:
Garry Tanandcapy-ai[bot] authored and GitHub committed 2026-11-18 15:48:59 +00:00
1 parent 782f4595ed
commit fda9dea79f
11 files changed
+127 -27

No files matched your search

+12
View File
@@ -224,6 +224,18 @@ or name a model in your request to override it for the invocation, including
resumed consultations. See [eval defaults and overrides](../CONTRIBUTING.md#testing--evals)
for capture, judge, and benchmark model selection.
**Subagents on Codex.** Codex has no Agent tool, so the Codex skills start every
audit and reviewer subagent (/ship's coverage, plan and docs audits, the native
review voices) with Codex's `spawn_agent` and wait on `wait_agent` until it reports
a final status. Turn multi-agent on first: `multi_agent = true` under `[features]`
in `~/.codex/config.toml`. Without it the dispatch fails and the step takes its
documented failure path. Codex has no read-only Plan subagent type, so when the
plan reviews' outside voice is unavailable they report it unavailable instead of
starting a native fallback. /ship's review record still only counts a reviewer
whose brief gstack itself sent: a reviewer started some other way has no recorded
send, so `gstack-review-dispatch verify` reports its brief `NOT_SENT` and prints
`VERIFY: dirty` (exit 1) rather than trusting it.
## The design binary's models
The design binary (`$D`) uses OpenAI's `gpt-5.5` by default, with image