Changelog

0.1.13

• Clarified agent workflow contracts: routine validated commits on recognized feature branches remain autonomous after required review; push and later lifecycle actions stay separately gated.
• Simplified and re-aligned the shared directives around outcome, evidence, runtime authority, blind review, bounded repair, and autonomous routine work.

0.1.12 (2026-08-11)

• Streamlined canonical agent directives by centralizing universal contracts and adding bounded autonomy, work-unit budgets, scope control, process lifecycle evidence, and checkpoint action boundaries.
• Hardened the Hermes adapter by wiring native session and subagent lifecycle trust, immutable mode and child safety allowlists, and the current Hermes plugin registration contracts.

0.1.11 (2026-08-04)

Fixed

• Fixed the Hermes delegation rules adaptation in the synced global rules. The route-scoped rewrite of the canonical Delegation section had left a sync findAndReplace transform silently no-oping, so the generated global-rules/SKILL.md shipped the raw canonical wording instead of the Hermes adaptation. Hermes agents again get the correct guidance: on focused/full routes, use only the 7 maestria specialists and never delegate to Hermes’ built-in explore or general agents, which bypass the pipeline.

0.1.10 (2026-08-03)

Selective routing contract

The orchestrator now routes by task class across direct, focused, and full routes. Small explanations and tiny edits run directly or through one specialist; complex or high-risk work uses the full pipeline. fein selects the full route, sonar is research-only, and blitz is an explicit low-risk/direct bypass that does not waive safety floors. Where the host cannot enforce separate sessions, the maker/checker split is advisory and stated as such.

0.1.9 (2026-07-29)

Rule-override conditions and communication conventions

• Added 6 explicit override conditions to the orchestrator prompt (user skip request, safety, mode override, frustration escalation, rule conflicts, explanation requests).
• Two new shared rules: report errors matter-of-factly and lead with the action.
• Review triage now chains directly into the commit flow after approval.

0.1.8 (2026-07-28)

Blind review and fail-loud iteration exit

• The reviewer no longer receives the builder’s handoff notes or self-assessment; it evaluates only the diff, requirements, and acceptance criteria.
• When the review loop runs 3 cycles with unresolved issues, the pipeline stops and escalates with a structured report requiring an explicit override to continue.

0.1.7 (2026-07-22)

Directive recomposition

• Restructured core prompts with clearer sections and emphasis on critical rules.
• Added structured handoff verification checklists to all specialist agents.
• Standardized “before reporting done” completion checks across all agents.
• Added a parallelization table and multi-lens review swarm capability.
• Made prompt instructions platform-agnostic for consistent behavior across platforms.

0.1.6 (2026-07-21)

• Sync workflow mode commands (fein/sonar/blitz) through canonical source pipeline.

0.1.5 (2026-07-21)

Compact agent directives

Agent directives compressed to cut orchestrator context usage by ~38% without removing rules or intent.

0.1.4 (2026-07-17)

Changed

• Orchestrator behavior: Simple tasks now run directly instead of always being delegated to subagents. Complex tasks (multi-file, cross-domain, risky) still use specialist routing. Routine changes are faster with less overhead - no action needed on your part.

0.1.3 (2026-07-17)

Added

• pre_gateway_dispatch hook: Intercepts /fein, /sonar, /blitz, /mode, /review, /plan commands before the agent-busy check
• Role-based permission enforcement (historical implementation): [MAESTRIA_ROLE: <specialist>] in delegate_task context enforced per-specialist tool restrictions
• Transform hook: Write operations in fein/blitz mode append methodology annotation to tool results

Fixed

• Telegram forum topic responses now route to the correct thread instead of General

0.1.2 (2026-07-16)

Fixed

• Plugin loading failure due to src/ layout - added root __init__.py shim

0.1.1 (2026-07-16)

Initial release - same as 0.1.0 (versioning alignment).

0.1.0 (2026-07-16)

Added

• Plugin scaffold with plugin.yaml + register(ctx) entry point
• 9 skills: 7 specialists + orchestrator + global rules
• Mode system with file persistence (/fein, /sonar, /blitz, /mode)
• pre_llm_call hook for mode injection into user message
• pre_tool_call hook for mode-based tool gating (sonar write-block)
• llm_execution middleware (opt-in mode footer via MAESTRIA_MODE_FOOTER=1)
• 7 slash commands: /fein, /sonar, /blitz, /mode, /mode-clear, /review, /plan
• OpenCode CLI routing tool (opencode_route) for delegating complex coding tasks
• Subagent lifecycle hooks (subagent_start/stop for pipeline tracking)
• Session lifecycle hooks (on_session_start/end)
• transform_tool_result hook (observer-mode write logging)
• Distribution via hermes plugins install from this repository (git-based)

Changed (post-release design doc alignment)

• Memory engine agnosticism - removed stale Mnemosyne integration claim, removed memory/kanban probes
• Dead code removed: memory.py (MemoryManager.record() never called)
• Cross-platform detection: shutil.which() replaces subprocess.run(["which", ...])
• Plugin check: node -e "require(...)" replaces fragile npm filesystem walker