This release adds 3 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Summary
AI summaryAdded question mode for read‑only answers triggered by markers ending with ?, starting with ? or using rik?:
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Medium |
Adds question mode for read-only answers from markers with ? syntax Adds question mode for read-only answers from markers with ? syntax Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Feature | Medium |
Adds dynamic command tools defined per file using `rik +tool:` syntax Adds dynamic command tools defined per file using `rik +tool:` syntax Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Feature | Medium |
Adds in‑memory question tracking for watch mode Adds in‑memory question tracking for watch mode Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Feature | Medium |
Adds `marker_limits_edition_range` guardrail to keep edits scoped to one marker region Adds `marker_limits_edition_range` guardrail to keep edits scoped to one marker region Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Dependency | Medium |
Depends directly on `rig-core` to remove unused integration dependencies Depends directly on `rig-core` to remove unused integration dependencies Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Performance | Medium |
Improves dynamic‑tool, denied‑request, and tool‑parameter output handling Improves dynamic‑tool, denied‑request, and tool‑parameter output handling Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Bugfix | Medium |
Fixes reverting of completed work when the agent uses `send_message` Fixes reverting of completed work when the agent uses `send_message` Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Bugfix | Medium |
Fixes continuation of scanning immediately after a multi‑line marker Fixes continuation of scanning immediately after a multi‑line marker Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Refactor | Low |
Changes task processing to handle markers individually and keep question output focused Changes task processing to handle markers individually and keep question output focused Source: llm_adapter@2026-06-10 Confidence: high |
— |
| Refactor | Low |
Makes halt markers local to their respective marker Makes halt markers local to their respective marker Source: llm_adapter@2026-06-10 Confidence: high |
— |
Full changelog
rik can answer questions now.
Added
- Question mode for read-only answers from markers ending with
?, starting with?, or usingrik?: - Dynamic command tools defined per file with
rik +tool: <command> - In-memory question tracking for watch mode
marker_limits_edition_rangeguardrail for keeping edits scoped to one marker region
Changed
- Process task markers individually and keep question output focused on the answer
- Make halt markers local to their marker
- Improve dynamic-tool, denied-request, and tool-parameter output
- Depend directly on
rig-coreto remove unused integration dependencies
Fixed
- Avoid reverting completed work when the agent uses
send_message - Continue scanning immediately after a multi-line marker
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About rik
All releases →Beta — feedback welcome: [email protected]