30 lines
1.4 KiB
Org Mode
30 lines
1.4 KiB
Org Mode
#+TITLE: PRD: PSF Core Role Automation
|
|
#+STATUS: FROZEN
|
|
|
|
* 1. Purpose
|
|
Define the automated behaviors for the specialized agents in the PSF Consensus Loop.
|
|
|
|
* 2. User Needs
|
|
|
|
** 2.1 Autonomous Role Progression
|
|
As a user, I want the system to handle the "Baton Pass" between agents.
|
|
- When I finish a PRD and mark it `FROZEN`, the Architect should wake up without my intervention.
|
|
- When the Architect signs a Protocol, the Analyst should automatically start drafting the tests.
|
|
|
|
** 2.2 Semantic Enforcement
|
|
I need the agents to act as "Safety Gates."
|
|
- The Coder must be physically blocked (via Lisp check) from implementation if the `tests/` directory is empty or if the `PROTOCOL.org` is in `DRAFT`.
|
|
|
|
** 2.3 Knowledge Provenance
|
|
I need the Scribe-RCA to link failures back to specific commits and sessions, ensuring that my Institutional Memory is a high-fidelity record of the system's evolution.
|
|
|
|
** 2.4 Dynamic Chaos Testing
|
|
I need a QA Specialist that proactively breaks things.
|
|
- The system must have a "Chaos Gauntlet" that runs against projects in the :BUILD or :CHAOS state.
|
|
- It must perform fuzzing, dependency sabotage, and persistence testing (killing the Lisp image).
|
|
|
|
* 3. Success Criteria
|
|
...
|
|
- [ ] **Chaos Trigger:** Agent perceives project in :BUILD state and triggers the Chaos Gauntlet.
|
|
- [ ] **Sabotage Simulation:** Chaos Specialist can successfully inject a failure and verify the system's recovery.
|