The hyperplan trigger \b(hyperplan|hpp)\b/i matched 'hpp' inside common C++ header references like 'check interface.hpp' or 'open buffer.hpp'. The leading '.' is a non-word character, so \b is already satisfied and the false positive fires the hyperplan-mode prompt on routine code questions.
Split the alternation so 'hpp' additionally requires that the preceding character is neither a word character nor a '.'. This preserves every existing trigger ('hpp do this', '/hpp ...', mid-sentence usage, mixed case) while rejecting filename uses of the .hpp extension. The longer 'hyperplan' keyword keeps the original \b boundary semantics.
Reproduction (added regression tests):
- 'please help to check interface.hpp' must NOT fire
- 'open src/include/audio/buffer.hpp and fix the leak' must NOT fire
All 14 cases in hyperplan.test.ts pass (12 existing + 2 new), broader keyword-detector suite stays green (92 pass), typecheck clean.
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>
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>
- 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)
- P1: When allowedEnvVars is provided, PATH was missing from the base
restricted env, causing non-builtin commands to fail at exec time
- P2: Allowlisted HOME/CLAUDE_PROJECT_DIR could overwrite normalized
values from getHomeDirectory()/cwd with ambient process.env values
- P2: Test suite mutated shared pluginHooksState singleton without
resetting it in afterEach, causing cross-test state leaks
- Apply mcp_env_allowlist to plugin hooks: intersect HTTP allowedEnvVars
with MCP allowlist, set command allowedEnvVars to full MCP allowlist
- Scrub process.env in executeHookCommand when allowedEnvVars provided
- Add PluginHooksState class with per-directory Map storage
- Add PluginHooksConfig interface for typed boundary layer
- Pass directory context through hook-config-handler
- Add 16 tests across 4 files (40 assertions) covering allowlist
filtering, env scrubbing, directory isolation, and edge cases
- Remove unnecessary 'as' type assertions, use discriminated union
narrowing instead
Todo continuation could reinject with a lowercase built-in config key such as hephaestus when no registered alias was available. OpenCode prompt dispatch expects the prompt-facing agent name, so the continuation failed with an agent-not-found error.
Normalize the dispatch fallback through the existing prompt agent display-name resolver and pin the lowercase Hephaestus regression.
Gate idle ready notifications on the existing background-task continuation marker so cmux does not receive premature ready alerts while delegated work is still active.
Constraint: Reuse marker state from background task lifecycle without adding new notification config
Rejected: Patch cmux directly | notification readiness belongs upstream in OMO
Confidence: high
Scope-risk: narrow
- Drop the back-import of src/shared/logger so @oh-my-opencode/rules-core
stays free of host-adapter dependencies (ROADMAP package layering
invariant). Expose setSisyphusRuleDeprecationLogger(logger) for hosts to
inject their own logger; default is a noop.
- Wire the host injection in src/hooks/rules-injector/rule-file-finder.ts
as a module-level side effect so existing behavior is preserved.
- Add realpath boundary check to findRuleFilesRecursive and
validFileRealPath. Project rule scanners now refuse entries whose
realpath escapes the rule-source root, closing the symlink-escape
vector where a malicious repo could point .github/copilot-instructions.md
(or any .omo/rules/* entry) at ~/.ssh/id_rsa and have the rule injector
pull the secret into model context.
Accept apply_patch edits from nested result/metadata file lists and from raw patchText args when OpenCode does not provide direct metadata.files.
Forward tool args through the after-hook pipeline so comment-checker can inspect raw apply_patch inputs while preserving existing write, edit, and multiedit routing.
Tests: bun test src/hooks/comment-checker/hook.apply-patch.test.ts src/hooks/comment-checker/hook.before-after.test.ts src/plugin/tool-execute-after.test.ts
Tests: bun test src/hooks/comment-checker
Tests: bun run typecheck
Tests: bun run build
The full Bun suite can remove the shared rules-injector storage directory between a parent-directory check and the file write. Save operations now create the directory immediately before writing and retry once if ENOENT still wins the race.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Ports the codex-rules transcript-aware dedup strategy: when a session cache is fresh (process restart or compaction-cleared cache), the injector now scans prior tool outputs for the `[Rule: <relativePath>]` banner and pre-populates the cache so duplicate rule injections are suppressed even when the persistent JSON has been lost. Hydration runs at most once per session per process, fails open on transport errors, and is short-circuited when the same banner reappears.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Reduce the ralph loop continuation prompt from ~15 lines to ~3 lines:
- Keep system directive prefix for proper identification
- Keep the essential 'output <promise>DONE</promise>' instruction
- Keep the original task prompt (needed context)
- Remove verbose boilerplate about reviewing progress and not stopping
When default_mode.ultrawork is enabled, inject the ultrawork behavioral
instructions into the system prompt (experimental.chat.system.transform)
instead of prepending them to the user's visible chat message.
- system-transform.ts: No-op handler now injects ultrawork message into
output.system[] when defaultMode.ultrawork is active. Checks for existing
<ultrawork-mode> tag to avoid re-injection after compaction.
- keyword-detector/hook.ts: Removes the visible text injection
(output.parts[].text prepend) for default_mode.ultrawork path. Only
shows the toast as the user-facing indicator.
- plugin-interface.ts: Wires getUltraworkMessage from keyword-detector
into createSystemTransformHandler.
Result: Ultrawork mode activates silently via system prompt — the chat
stays clean, and only a toast shows 'Ultrawork Mode Active'.
Add new `default_mode` config section with two boolean fields:
- `ultrawork`: Auto-inject ultrawork mode prompt on main session start
without requiring the "ultrawork"/"ulw" keyword. Wired through the
keyword-detector hook — injects once per session, respects existing
guards (non-OMO agents, planner agents, subagent sessions).
- `ralph_loop`: Auto-start ralph loop on first main session message
without requiring /ralph-loop or /ulw-loop commands. When ultrawork
is also enabled, the loop starts in ultrawork mode.
Usage:
```jsonc
{
"default_mode": {
"ultrawork": true, // Always get ultrawork prompt on start
"ralph_loop": true // Auto-start ralph loop
}
}
```
Files: 7 modified/added, ~65 LOC added.
MoerAI