Epic: Guarded actions, structured steering, and Cockpit chat

[smota]

Background & Problem Statement

Epic: #119

Cockpit should allow safe operator interaction, but remote UI must not become hidden authority or bypass AgentFlow gates.

Proposed Solution

Implement structured action gateway and chat/steering lane. Chat creates typed intents and GitHub-visible durable evidence when material.

Requirements

• Typed intents: ask-status, add-clarification, request-checkpoint, request-human-review, draft-follow-up, update-issue-section, post-handover, start-next-phase.
• Action preview and confirmation for GitHub writes.
• All material decisions become GitHub comments/body updates using issue standards.
• High-assurance gates cannot be bypassed; validation/review cannot be marked passed by chat.
• Executor target changes require resolver evidence.
• Audit log for every action.

Acceptance criteria

• Chat status answers cite GitHub/role-pass evidence.
• User can draft follow-up issue from review finding.
• User can post handover or clarification via guarded action.
• Forbidden actions are blocked with policy reason.
• Material steering produces durable GitHub evidence.

Test plan

• Intent parser tests.
• Guarded action authorization/confirmation tests.
• Policy-block tests for gate bypass and merge attempts.
• E2E chat-to-comment flow.

Workflow classification

High-assurance remote write surface. Requires security review, tester, and tech writer.

[github-actions]

Integrated into development by PR #126: #126

Merge commit: a579156d0b0568f70dfcf3691c1ae8eff164ee22
Release/promotion to main is tracked separately.