@caupulican/pi-adaptative 0.79.0 → 0.80.1

package/CHANGELOG.md

@@ -2,6 +2,23 @@
 
 ## [Unreleased]
 
+## [0.80.1] - 2026-06-01
+
+### Changed
+
+- Enabled grouped interactive tool panels by default for all tools, while preserving explicit cross-tool grouping and blank-string opt-out.
+
+## [0.80.0] - 2026-06-01
+
+### Added
+
+- Added low-config autonomy presets via `autonomy.mode` and `/autonomy status|off|safe|balanced|full`; `full` now means standing autonomy with post-turn reflection whenever concurrency allows, plus grants for memory, skills, user/project extensions/tools, autonomy tuning, and authorized self-modification source edits while keeping publish/push/tag/release/credential/destructive actions foreground-approval gated.
+- Added Auto Learn reflection review: when Auto Learn/autonomy is enabled, Pi can launch a bounded background learner after corrective or complex tool-heavy turns, with separate cooldown/status reporting and authority instructions derived from `autonomy.mode`.
+
+### Security
+
+- Added prompt-injection/exfiltration scanning for eagerly injected `AGENTS.md`, `CLAUDE.md`, and `GEMINI.md` context files; suspicious files are replaced with a blocked notice instead of entering the system prompt.
+
 ## [0.79.0] - 2026-06-01
 
 ### Added

package/dist/core/agent-session.d.ts

@@ -321,6 +321,7 @@ export declare class AgentSession {
     private _normalizePromptSnippet;
     private _normalizePromptGuidelines;
     private _buildSelfModificationPrompt;
+    private _buildAutonomyPrompt;
     private _rebuildSystemPrompt;
     private _runAgentPrompt;
     private _handlePostAgentRun;