1 changed files with 50 additions and 0 deletions
@@ -110,6 +110,24 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
  `docs/development/automation-tracking.md` and the new
  `docs/development/docs-writer.md` reference.

+- **Dual-Account PR Review Architecture**: Introduced a dedicated reviewer account
+  (`FORGEJO_REVIEWER_*` credentials) separate from the primary bot account. The
+  `product-builder` passes reviewer-only credentials through the entire PR review
+  chain (`pr-review-pool-supervisor` → `pr-reviewer`), preventing self-approvals
+  where the same account both creates and approves PRs. All review operations in
+  `pr-reviewer` now use `curl` exclusively against the Forgejo REST API, bypassing
+  MCP tools that would attribute actions to the wrong account.
+
+- **Agent System Specification — Section 22 (Division of Responsibilities)**: Added
+  comprehensive responsibility-first analysis covering 144 named responsibilities
+  (R-01 through R-144) organized into 8 categories with 29 sub-categories and complete
+  bidirectional cross-references. Includes Section 22.0 (Categorized Responsibility
+  Index), Section 7.9 (Universal Supervisor Responsibilities — 13 shared baseline),
+  Section 22.145 (Redundancy Analysis Summary), and 30 agent back-reference tables.
+  Categories: A (Development Lifecycle), B (Quality Assurance), C (Ticket Hygiene),
+  D (Architecture), E (Operational Health), F (Human Interaction), G (Strategic
+  Governance), H (Project Standards).
+
 ### Changed

 - **Decision Tree Full ULID Display** (#5825): The `agents plan tree` command now
@@ -141,6 +159,29 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
  `AUTO-INF-POOL`, `AUTO-ARCH`, `AUTO-EPIC`, `AUTO-EVLV`, `AUTO-GUARD`, `AUTO-SPEC`,
  `AUTO-TIME`, `AUTO-PROJ-OWN`, and `AUTO-PROD-BLDR`.

+- **ACMS Context Hydration**: Fixed ACMS indexing pipeline not wired into CLI —
+  `ContextTierService` started empty on every CLI invocation so LLM received zero file
+  context during plan execution. Added `context_tier_hydrator.py` that reads files from
+  linked project resources (via `git ls-files` or `os.walk`) and stores them as
+  `TieredFragment` objects in the tier service. Hydration runs automatically before context
+  assembly in `LLMExecuteActor.execute()`. Respects max file size (256KB), total budget
+  (10MB), binary file exclusion, and `.git`/`node_modules`/`__pycache__` directory
+  skipping. (#1028)
+
+- **ATM Estimated Cycle Interval Automation**: The `automation-tracking-manager` now
+  automatically calculates and injects `**Estimated Cycle Interval**: Nmin` into every
+  tracking issue body when agents pass `sleep_interval_default`. Uses a rolling average
+  formula (`new = old * 0.90 + actual * 0.10`) based on the time elapsed since the
+  previous tracking issue. Agents no longer need to calculate or track this value
+  themselves. See `docs/development/automation-tracking.md` for the full protocol.
+
+- **Tracking Startup State Recovery Protocol**: All 17 supervisor agents now follow the
+  mandatory READ-before-CREATE startup sequence: call `READ_TRACKING_STATE` first to
+  recover previous session state (cycle number, offline duration, last activity), then
+  call `CREATE_TRACKING_ISSUE` to open the new cycle. This prevents agents from losing
+  context after crashes or restarts and enables informed recovery decisions based on
+  offline duration.
+
 ### Fixed

 - **Plan Concurrency Race Condition** (#7989): Fixed critical race condition in `execute_plan()` and
@@ -203,6 +244,15 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
  `pr-merge-pool-supervisor` to the product-builder's supervisor launch list (18 total
  supervisors). Updated all numeric references, pre-flight checklists, and validation logic.

+- **`pr-reviewer` MCP Token Attribution**: The `pr-reviewer` agent now uses `curl`
+  exclusively for all Forgejo API operations (PR details, diffs, CI status, posting
+  reviews, posting comments). This prevents review actions from being attributed to the
+  primary bot account when the reviewer account credentials are in use, fixing the
+  self-approval problem where a bot would approve its own PRs.
+
+- **`pr-ci-test-fixer` Stale Agent References**: Removed stale references to the old
+  `pr-self-reviewer` agent name. All references now correctly point to `pr-reviewer`.
+
 ---

 ## [3.8.0] — 2026-04-05