Context
The ticket-create skill currently enforces a "Visible Proposal Protocol" (Section 6). This protocol mandates that agents must output the proposed ticket content in the chat to the human commander before (or alongside) invoking the create_issue MCP tool. Historically, this served as an early-intervention gate to prevent poorly scoped tickets from polluting the tracker.
The Problem
This human-in-the-loop gate acts as an artificial bottleneck for fully autonomous agent workflows. It conflicts with the project's evolving philosophical paradigm ("Tobi is the merge gate"), which centralizes human oversight at the final Pull Request squash-merge stage rather than micromanaging upstream artifact creation.
The Architectural Reality
By enforcing a visible proposal for every ticket, we effectively create a "human every gate" scenario. If an autonomous agent (like a background tech-debt radar or retrospective daemon) wants to log a ticket, it shouldn't need to halt the swarm for chat-based approval. The MCP server and downstream PR review provide sufficient safety nets.
The Fix
Update .agent/skills/ticket-create/references/ticket-create-workflow.md to completely strip Section 6 (Visible Proposal Protocol). Agents should be instructed to synthesize the Fat Ticket and immediately execute the create_issue tool without seeking explicit user-side verification in the chat output.
Acceptance Criteria
Out of Scope
Relaxing the Human-in-the-Loop Protocol for PR merges. The final merge remains strictly manual.
Avoided Traps / Gold Standards Rejected
- Trap: Maintaining legacy "human every gate" bottlenecks out of fear of tracker pollution. The five-stage challenge chain (Gate 1-5) inside the prompt is sufficient to prevent low-quality tickets; the human does not need to verify the output manually before creation.
Origin Session ID: 0b29a8fa-c6b0-42e2-ab3b-8015a99db2d8
Context The
ticket-createskill currently enforces a "Visible Proposal Protocol" (Section 6). This protocol mandates that agents must output the proposed ticket content in the chat to the human commander before (or alongside) invoking thecreate_issueMCP tool. Historically, this served as an early-intervention gate to prevent poorly scoped tickets from polluting the tracker.The Problem This human-in-the-loop gate acts as an artificial bottleneck for fully autonomous agent workflows. It conflicts with the project's evolving philosophical paradigm ("Tobi is the merge gate"), which centralizes human oversight at the final Pull Request squash-merge stage rather than micromanaging upstream artifact creation.
The Architectural Reality By enforcing a visible proposal for every ticket, we effectively create a "human every gate" scenario. If an autonomous agent (like a background tech-debt radar or retrospective daemon) wants to log a ticket, it shouldn't need to halt the swarm for chat-based approval. The MCP server and downstream PR review provide sufficient safety nets.
The Fix Update
.agent/skills/ticket-create/references/ticket-create-workflow.mdto completely strip Section 6 (Visible Proposal Protocol). Agents should be instructed to synthesize the Fat Ticket and immediately execute thecreate_issuetool without seeking explicit user-side verification in the chat output.Acceptance Criteria
ticket-create-workflow.md.ticket-proposal-template.mdfrom theassetsfolder if it is no longer referenced.AGENTS.mdor other skills that cite the visible proposal requirement.Out of Scope Relaxing the Human-in-the-Loop Protocol for PR merges. The final merge remains strictly manual.
Avoided Traps / Gold Standards Rejected
Origin Session ID: 0b29a8fa-c6b0-42e2-ab3b-8015a99db2d8