The test file modifies process.exitCode and emits process signals which can
leak into bun test's exit code. Add:
1. mock.module() sentinel to route to isolated batch (following abort-with-timeout.test.ts pattern)
2. Global afterAll() hook that resets process.exitCode = 0 before test runner checks it
3. Remove direct checks of process.exitCode in assertions - only check that process.exit() was called with the right code via spy
This ensures bun test exits with code 0 even after tests verify process.exit behavior.
Fixes#3792
Property 'parentSessionID' does not exist on type 'LaunchInput' / 'BackgroundTask'.
The correct casing is 'parentSessionId' (camelCase with lowercase 'd').
Fixes CI build failure on dev branch.
When loading Claude Code plugins, the hooks JSON config already had
${CLAUDE_PLUGIN_ROOT} replaced via resolvePluginPath. However, command
markdown bodies, agent prompts, and skill content were passed through
as-is, leaving the variable unresolved. This caused plugin commands
(e.g. codex:review) to fail when executed in OpenCode because the shell
had no CLAUDE_PLUGIN_ROOT environment variable set.
Apply resolvePluginPath to:
- command-loader.ts: command body before wrapping in template
- agent-loader.ts: agent prompt body
- skill-loader.ts: skill body after @path resolution
The global test-setup.ts (preloaded via bunfig.toml) already snapshots
process.env in beforeEach and restores it in afterEach for every test.
Per-file env tracking is duplicate work and creates four different
patterns for the same problem.
Affected files:
- src/shared/claude-config-dir.test.ts (beforeEach/afterEach pair)
- src/shared/plugin-command-discovery.test.ts (ENV_KEYS + envSnapshot)
- src/features/claude-code-agent-loader/loader.test.ts (try/finally)
- src/features/skill-mcp-manager/connection-env-vars.test.ts (ORIGINAL_ENV)
isRetryableModelError() now checks the HTTP status code (429/503/529)
in addition to existing message pattern matching. This ensures rate
limit errors trigger model fallback regardless of error message format
or language (e.g., Chinese GLM errors).
Changes:
- ErrorInfo interface extended with statusCode?: number
- isRetryableModelError() checks statusCode after STOP patterns, before
message pattern fallback
- extractErrorStatusCode() added to error-classifier.ts (supports
statusCode, status, code, response.status fields)
- GLM-specific STOP patterns added: daily call limit, in arrears,
fair use policy, recharge and try — these prevent quota/billing 429s
from being treated as transient rate limits
- statusCode propagated through tryFallbackRetry and manager.ts
400 intentionally excluded from statusCode check (permanent client error).
Atlas continuation only treated running background tasks as active work. When a delegated subagent had been launched but was still waiting for session creation, the task remained pending and Atlas could inject another continuation too early.
Treat pending tasks as active background work in the continuation injector and add regression coverage for the pending-session-creation race so delegated work is allowed to acquire a session before Atlas resumes the plan.
Tests: bun test src/hooks/atlas/boulder-continuation-injector.test.ts src/hooks/atlas/index.test.ts; bun run typecheck
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Rename BackgroundTask and attempt ID fields to camelCase across background-agent consumers while moving BackgroundManager construction to a single config object.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
AgentOverridesSchema silently strips custom agent keys during Zod
parsing because only 14 built-in names are explicitly defined. Add
.catchall(AgentOverrideConfigSchema.optional()) so user-defined agent
configs survive validation and reach downstream consumers like
resolveModelAndFallbackChain().
Fixes#3229.
Bare "403" and "forbidden" substring patterns (added in 034744cb to
retry the "Selected provider is forbidden" case from PR #3706) matched
any error message containing those tokens — tool-level 403s, file-perm
"forbidden" messages, unrelated upstream errors that happened to spell
"forbidden". The legacy model-fallback path then armed setPendingModelFallback
on those unrelated errors, jumping Sisyphus to claude-opus-4-7 (first
entry of its fallback chain) regardless of the user's configured model.
Replace the bare patterns with the specific phrases PR #3706 actually
targeted, and add regression tests asserting unrelated 403/forbidden
messages stay non-retryable.
Reported-by: ilove_borshch on Discord (#omo-help)
- Add Manual QA Gate as a non-negotiable surface-tool mapping
(interactive_bash for TUI, playwright for browser, curl for HTTP,
driver script for library) to Hephaestus, Sisyphus-Junior, and
Sisyphus' direct-execution mode.
- Restore the dig-deeper trio (tool persistence / dig deeper /
dependency checks) as orthogonal paragraphs so each carries its
own cognitive trigger instead of a fused single statement.
- Harden investigate-before-acting from a soft phrase to a dedicated
block: never speculate about unread code, re-read on every
task hand-off, the worktree may have changed.
- Add 'Parallelize aggressively' as its own block. Reads, searches,
diagnostics, and background sub-agents all batch into a single
response by default.
- Add 'No defensive code, no speculative legacy' to discourage
speculative backward-compatibility branches and unrequested
defensive validation.
- Absorb review-intent and frontend anti-slop coverage so the
prompt stays self-sufficient when the omo agent prompt replaces
the provider prompt.
- Replace literal apply_patch instructions with GPT_APPLY_PATCH_GUIDANCE
(use edit/write tools) so the prompt no longer contradicts the
apply_patch deny that the agent permission applies on GPT models.
- Sisyphus-Junior gains a Review tasks block and a default-behavior
fallback for when the runtime category context is missing or sparse.
- Sisyphus gains an explicit Hard invariants block listing type-
suppression bans, destructive-git bans, and Oracle-completion gating.
- Restore dynamic injections that round out the orchestrator/worker
context: category+skills delegation guide, delegation table,
Oracle dynamic guidance, key triggers, non-Claude planner
reminder. Hephaestus regains optional category delegation while
keeping direct execution as the default.
- Drop em dashes; search guidance points at rg directly throughout.
When shutdown() itself emitted uncaughtException (e.g. EPIPE while closing
a broken pipe), the error listener re-entered itself, re-logged, re-ran
cleanup, and threw EPIPE again. The 6 s forced-exit timer could not fire
because every re-entry stalled the event loop with fresh synchronous work.
Users hit this after v3.17.5 and observed 100+ GB of log lines written to
disk within minutes, with one confirmed report of a 157 GB log file filling
the filesystem.
Detaching the listener with process.off() before running log() + handler()
breaks the loop at the first re-emit: the second event has no listener to
invoke, and the first invocation's scheduleForcedExit() proceeds normally.
YeonGyu-Kim