Cubic caught that clearing recent parent activity on session.idle could reopen the stale-idle race. Keep the fresh activity marker until its normal expiry so background completions still defer while the parent turn is actively reasoning.
Red: parent idle after fresh reasoning delta dispatched an unexpected parent wake. Green: focused parent-wake regression suite passes.
Record fresh parent session message activity before parent-wake flushing so stale idle status cannot dispatch a background completion into a live reasoning turn.
Add regression coverage for the Discord 4.2.3/OpenCode 1.15.5 duplicate-branch repro shape where a parent reasoning delta arrives before the background all-complete wake.
Refs #4212
Refs #4019
Refs #3774
Plan: plans/background-notification-active-turn-queue.md
When the working directory contains a .git folder the OpenCode server
normalises the project root to the git root before persisting messages.
This creates a race: the 429/503/529 error event can fire before the
user's message is committed to storage, so session.messages returns []
and getLastUserRetryParts returns an empty array. The previous code
treated that as a silent no-op (cleared all retry state, Sisyphus stalled).
Fix: when fetchedParts is empty, emit a structured log explaining the
.git-directory race and fall back to a synthetic { type:"text", text:"continue" }
part — matching the pattern already used by autoContinueAfterFallback in
event.ts. The fallback dispatch always proceeds regardless of whether
the messages API can return user parts.
Update four tests that fired two consecutive session.error events relying
on the old silent-stop behaviour: add top-level model fields to the second
error so the awaiting-fallback gate recognises it as coming from the
dispatched fallback model and lets it through normally.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
On Windows, path.resolve() returns paths with backslash separators. The previous endsWith("dist/cli.js") check uses forward slashes and always returned false on Windows, causing both LSP and ast-grep MCPs to fall through to the bootstrap path even when the dist cli exists. Result: LSP MCP completely unusable on Windows with MODULE_NOT_FOUND for dist/packages/lsp-tools-mcp/dist/cli.js.
Fix: derive a platform-aware suffix at module load time by replacing forward slashes in DIST_CLI_REL / SOURCE_CLI_REL with path.sep, then use that suffix in the endsWith check.
Verification: all 4 src/mcp/lsp.test.ts cases pass on Windows (previously 2 failed); all 13 src/mcp/ast-grep.test.ts cases pass (previously 2 failed). Total: 17/17 src/mcp tests green. bun run typecheck clean.
Maintainer feedback (#4071 review): the original guard rejected
sisyphus and atlas as subagent targets even from team-mode where
resolveMember() intentionally calls resolveSubagentExecution with
allowPrimaryAgentDelegation: true. Per AGENT_ELIGIBILITY_REGISTRY
(src/features/team-mode/types.ts), only prometheus is hard-reject;
sisyphus and atlas are explicitly verdict: 'eligible' for team
membership.
Shrink COORDINATOR_AGENT_NAMES to ['prometheus'] so the guard
aligns with the registry's authoritative classification, document
the scoping rule in a comment, and add regression tests covering:
- sisyphus is NOT blocked by the coordinator guard (registry eligible)
- atlas is NOT blocked by the coordinator guard (registry eligible)
- prometheus IS blocked even when allowPrimaryAgentDelegation: true
(registry hard-reject is authoritative)
Fixes the 5 zauc-mocks resolver tests that were locking in the
wrong rejection set (including 'allows delegating to a primary
agent when allowPrimaryAgentDelegation is enabled'). The one test
asserting the literal primary-agent error string for Prometheus
display-name was loosened to a regex that accepts either guard's
message, since prometheus is now caught by the coordinator path
which fires before the primary-agent lookup.
Agents could select coordinator/meta agents (Prometheus, Atlas,
Sisyphus/Ultraworker) as subagent targets via task() / delegation,
producing duplicate orchestration loops and conflicting team state.
This is the inverse of #3987 / #4065 — symmetric guard on the
delegation TARGET side, using the same AGENT_ELIGIBILITY_REGISTRY
classification.
Add a runtime guard at the delegation entry point that rejects
task() calls whose subagent_type resolves to an agent marked as
hard-reject / coordinator-only in the eligibility registry, with
an actionable error naming the agent. Regression test asserts a
prometheus-targeted delegation is rejected before any subagent
session spawns.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
PLANNING_CONSULT_WARNING was prepended to the prompt forwarded to
subagent LLMs via task(). Its leading bracket-enclosed marker
`[SYSTEM DIRECTIVE: OH-MY-OPENCODE - PROMETHEUS READ-ONLY]` is
exactly the indirect-prompt-injection signature that Azure OpenAI
Prompt Shield flags in user-role content; on GPT-5.4 through
Azure, the model returns "I'm sorry, but I cannot assist with
that request." before any planning work runs, making Prometheus
non-functional on Azure.
The bracket marker was designed for internal hook-to-hook
filtering, but PLANNING_CONSULT_WARNING leaks it to external LLM
payloads. Replace the header with a neutral XML-tag wrapper
(`<planning-context source="prometheus-read-only">`) that Azure's
filter does not match while preserving the human-readable warning
body. Internal isSystemDirective() consumers are unaffected.
Regression test asserts the post-hook task() prompt does not
begin with the flagged bracket sequence.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replace 'FN.' shorthand with explicit examples 'F1.', 'F2.' to prevent
LLMs from generating literal 'FN.' labels that the parser rejects.
Identified by cubic.
- Strengthen Prometheus plan template with FORMAT constraint
- Add task label format check to Oracle phase-2 (N/6 → N/7)
- Add format checks to self-review checklist
- New plan-format-validator hook: compares raw checkbox count
against getPlanProgress() after plan writes, warns agent when
labels are malformed (0/0 or partial skip scenarios)
YeonGyu-Kim