Wide wishes made Claude Code write more than 16 MiB of stream-json and
stopped at "executor failed". The limit is now maxOutputBytes, 256 MiB
by default.
Co-Authored-By: Claude Haiku 4.5 (1M context) <noreply@anthropic.com>
- cluster-poc.mjs: image (in-cluster Kaniko), bootstrap, secrets, apply,
migrate --rehearsal|--final, public HTTPS entry, backup and restore
- One restricted Pod (platform, executor, Runner, PostgreSQL, Temporal,
MinIO, Caddy, backup) listening only on loopback except Caddy
- Protected checks run in per-workspace check Pods; cancellation during
start-up now always removes them
- Daily snapshots on the node (consistent dump with row counts, Temporal
online backup), copied to the Mac by launchd with SHA-256 checks and a
notification on failure; restore verified on the rehearsal copy
- Downloads and uploads survive the Mac's lossy link to the cluster
- Local public entry, evidence isolation labels, acceptance records
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With WF_CLAUDE_COMMAND_SANDBOX=true, Claude Code also gets Bash, run in
its built-in sandbox through --settings:
- The sandbox must start (failIfUnavailable), and no command may run
outside it (allowUnsandboxedCommands: false).
- No network: an empty allowedDomains with strictAllowlist, since an
absent allowlist leaves the network open. Local binding stays off,
because it would also open every local port, Temporal's included.
- Commands cannot read the home directories, the CLI's home or the
worker's own directory, except the workspace, a per-run scratch
directory and WF_CLAUDE_COMMAND_READ_PATHS; they write only to the
workspace and the scratch directory; credential variables are unset.
- WF_CLAUDE_COMMAND_ENV gives commands toolchain variables, such as the
POC's read-only Go module cache.
The executor probes the sandbox first (Seatbelt on macOS; bubblewrap
and socat on Linux) and keeps file tools only where it cannot start,
saying so on the timeline. The adapter now writes the execution
contract into the prompt, for the tools the agent actually has.
scripts/check-claude-command-sandbox.ts checks a machine with one short
real session and judges each probe by its effect, not the agent's
account. On the local Mac every probe was blocked; the login keychain
remains reachable, because Claude Code's Seatbelt profile always allows
the security server. That limit is documented.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude Code with a tenant key started with bypassPermissions and every
tool, Bash included, and loaded the repository's settings, hooks and MCP
servers. It now runs like the native development login already did:
restricted mode with dontAsk, only Read/Edit/Write/Glob/Grep, confined to
the workspace, and no repository or user settings. Every Claude Code run
is told the verifier runs installs and tests.
CodeBuddy's agent runs commands with no sandbox of its own. Like Codex
without its sandbox, it now starts only in a development (host) sandbox,
says so on the timeline, and stops with PROVIDER_SANDBOX_UNAVAILABLE on
Kata.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The first runs on k3s found these problems; a wish now goes from submission
through the Codex executor, independent verification and integration in
per-run pods.
- Development deployments refused to start because the images set
NODE_ENV=production. The local manifest and the sandbox Jobs now set
NODE_ENV to the deployment mode.
- The model proxy treated x-client-request-id as a per-request idempotency
key. Codex sends its conversation id there, so every second request of a
session failed with 409. Only Idempotency-Key names a single request now.
- Codex's command sandbox (bubblewrap) cannot start in a runc pod. Every
command failed and the agent retried until the budget ran out. The
executor now probes the sandbox before starting Codex. A Kata sandbox that
cannot run it waits as PROVIDER_SANDBOX_UNAVAILABLE. A development (host)
sandbox runs Codex without it, and the timeline records that.
- The prompt asked agents to commit, but Codex keeps .git read-only and the
platform commits the checkpoint itself. Agents now leave their changes
uncommitted, and get the platform's Git identity if they commit anyway. A
resolved merge is recorded even when its tree equals HEAD.
- scripts/k3s-local.mjs:
- excludes the VM and cluster ranges from a host proxy (NO_PROXY);
- uses its own Docker client configuration, so a stale credential
helper does not break builds;
- recovers from an interrupted k3s install;
- reports when Colima switches the current kubectl or Docker context.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adding an executor touched the contracts, the API's availability check, the
worker and the console separately. Now it is an adapter plus one entry in
@wishforge/contracts' providerCatalog (name, executable variable, whether the
model proxy isolates its credentials) and one in the worker's
providerRegistry; the API, the console's options and MCP's submit_wish read
the catalog, and a test checks every entry is registered with the worker.
The CodeBuddy adapter already parsed Claude-Code-style stream-json, so it is
now a StreamJsonCliAdapter with a dialect per CLI. Claude Code runs headless
with --verbose and --permission-mode bypassPermissions inside the per-run
sandbox and receives only ANTHROPIC_* credentials (WF_CLAUDE_BIN). Like
CodeBuddy it is development-only until the model proxy supports it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Verification ran git inside the tree under test, kept no command output and
treated every infrastructure fault as a test failure; repairs went through an
executor lease the coordinator should not hold; and a project without delivery
could never succeed. This reworks the worker execution path and the control-plane
contracts it depends on.
- Verification exports the exact commit to a directory without .git, installs
dependencies (network allowed) and runs the protected commands under
bubblewrap in kata mode or directly in host mode; evidence records exit code,
duration, digest, a redacted output tail and the isolation used. Production
rejects host evidence. A failed installation waits as infrastructure.
- Waiting reasons are an enum with a redacted detail; dispatcher, GitHub and
model-proxy states map onto it, and the console explains each one with a way out.
- Repairs are requested by the coordinator through /internal/runs/:id/repair;
the control plane derives the failure summary from the verifier's evidence and
enforces the repair budget. Continuing after the budget is spent grants one more
round; a new requirement revision starts a fresh budget.
- Execution identities may only move a run within their own stages and only to
running or waiting; success comes from integration evidence (no delivery
configured) or deployment acceptance. The unused preview stage is gone.
- Integration fast-forwards when the target has not moved, merges and re-verifies
otherwise, retries when a push finds the target moved, and turns merge conflicts
into a repair round that merges the current target for the agent to resolve.
- Successful runs that needed a repair leave a project lesson; active lessons are
injected into later prompts, and admins can list, add, disable and delete them.
- Executor progress (messages, commands, file changes, plan updates) is stored as
normalized events; previously agent messages were recorded as the literal word
"output". Provider failures keep the work as a checkpoint and wait as
provider_error instead of failing the run.
- Workers declare WF_SANDBOX_MODE=host|kata (host refused in production),
heartbeat with retries until the lease could have expired, and retry safe
control-plane reads with backoff.
- tests/local-loop.test.ts runs the loop end to end with real Git, the in-process
control plane behind its internal routes and a scripted agent.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Drop the WorkBuddy Cloud executor stub: it always reported
"unverified" and could never run. WorkBuddy remains an MCP client for
submitting wishes. Providers are now codex and codebuddy; an unknown
provider waits with an explicit reason instead of a fake adapter.
- Remove internal routes nobody calls (/internal/releases/:id/begin,
/accept, /failure and /internal/workspaces/:id/provisioned) with
their domain methods, the never-consumed release.ready and
release.rollback outbox events, an unreachable idempotency branch,
dead helpers in runtime/deploy.ts and an unused RunnerBroker
dependency.
- Enable noUnusedLocals in the base and web TypeScript configs; the
runtime package now extends the base config.
- Add tests/tsconfig.json and include it in `npm run typecheck`. This
surfaced tests still calling the deleted begin/accept path (now
exercised through the Runner protocol) and five latent type errors.
- Web: provider capability labels match the API's keys and the wish
form disables unavailable providers.
- Delete the invalid example Runner Job manifest; stop tracking the
generated apps/web/next-env.d.ts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adds prettier 3.9.8 (exact) with a shared config (single quotes,
120 columns, trailing commas), .prettierignore, .editorconfig and the
`format` / `format:check` scripts, then formats all TypeScript,
JavaScript, CSS, JSON and YAML sources.
Purely mechanical: no behaviour change. Typecheck passes, the vitest
suite and the 74 IDE node tests produce the same results as before
(socket-binding tests still cannot run inside this sandbox).
Previously the longest source line was ~3000 characters; reviewing the
following semantic changes now produces readable diffs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Imports the previously untracked working tree as-is so that the review
changes that follow (requirements -> design -> implementation) can be
reviewed as diffs and reverted if needed. No content changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>