$ the-wire · showcase
Codex presets and inline model flags, credential masking for remote exec
By RepoJournal · Filed · About Agentic Coding · Composed from the cited sources · methodology
22 people shipped this
copyberry
@copyberry
3 cited
saoudrizwan
@saoudrizwan
2 cited
viyatb-oai
@viyatb-oai
1 cited
elberthc-byte
@elberthc-byte
1 cited
Cassian433
@Cassian433
1 cited
GitHub Actions
@actions-user
Hona
@Hona
vimtor
@vimtor
neriousy
@neriousy
rekram1-node
@rekram1-node
jesussamuel-byte
@jesussamuel-byte
villahernandez-coder
@villahernandez-coder
jvargassanchez-dot
@jvargassanchez-dot
Srinath Padmanabhan
@srithreepo
Engel Nyst
@enyst
simonrosenberg
@simonrosenberg
VascoSch92
@VascoSch92
TheRealSpencer
@TheRealSpencer
Mikołaj Kondratek
@mkondratek
BarreiroT
@BarreiroT
alexhancock
@alexhancock
lifeizhou-ap
@lifeizhou-ap
Codex turned its composer into a configurable surface with reusable presets and inline model/reasoning/speed flags, while its remote exec path stopped handing real credentials to Bash commands.
Enable credential masking for remote exec-server commands (#54309) openai/codex
Supported non-login Bash commands on Linux and macOS now receive credential placeholders instead of real secrets, with the executor-local proxy injecting credentials only for matching destinations, and mappings persist across process launches within the same executor session scoped by environment and conversation.
Add `/presets` for reusable prompts and slash commands openai/codex
A TUI picker gives you ten numbered preset slots for reusable prompts and slash commands, created from composer text or your current model, reasoning effort, and speed settings, persisted under `tui.presets` in your user configuration. Saved inputs run through the normal queued submission path and your composer draft survives.
Support inline model, reasoning, and speed selection in `/model` openai/codex
`/model [model] --reasoning <level> --speed <tier>` now updates the current session without touching saved defaults, with both space-separated and `=` flag values accepted; omit the model to retarget the current one. Bare `/model` still opens the picker, and unknown flags or unsupported levels and tiers are rejected before anything applies.
Add a keyboard shortcut menu for presets openai/codex
The `/presets` recall picker moved to a responsive menu above the composer: press `0` through `9` to run a saved preset or create one in an empty slot, and `n`, `s`, `e`, `d` to create, start from current settings, edit, or delete. The new `tui.keymap.chat.open_presets` binding defaults to an implicit `leader p` that yields to existing bindings, and the action appears in `/keymap` with a shortc...
fix: add /v1 to host-only OpenAI Compatible base URLs once, on VS Code and JetBrains cline/cline
The extension had stopped appending `/v1` to host-only OpenAI Compatible base URLs, which would have started returning 404s on the next release for the common vLLM, LM Studio, Ollama, and proxy setups; `/v1` is now added once, on both VS Code and JetBrains.
fix(openrouter): use the context length OpenRouter reports aaif-goose/goose
OpenRouter sessions ignored the context length the provider reports and fell back to the 128,000 default for any model outside the canonical catalog unless `GOOSE_CONTEXT_LIMIT` was set; the provider now implements the context-limit hook from `/api/v1/models`, parsed once and refetched hourly, with a failed refetch keeping the old list and retrying after 60 seconds.
fix(cli): resolve settings placeholders after loading .env files google-gemini/gemini-cli
Settings placeholders were expanded against `process.env` before `.env` files were loaded, so a value defined only there, such as `"Authorization": "Bearer $GITHUB_MCP_PAT"`, was sent to MCP servers as the literal string and came back `400 Bad Format Authorization Header`; resolution now happens after `.env` loading.
feat(desktop): remove the Cloud sessions settings toggle cline/cline
Desktop Cloud sessions no longer require the Settings > General toggle: the `code-cloud-agents` PostHog rollout flag is now the only gate, so cloud turns on for everyone that flag already targets.
Action items
