Epic: Cockpit Goal Story replay

[smota]

Background & Problem Statement

AgentFlow Cockpit needs a read-only replay capability for short goals. The replay must recreate how a goal moved from intent to outcome without rerunning agents or mutating GitHub. It should present AgentFlow evidence and decisions, not raw transcripts or generic event feeds.

Proposed Solution

Add Goal Story / Replay as an opinionated Cockpit view. It reconstructs a compact timeline from GitHub issues, comments, PRs, commits, checks, role-pass markers, handovers, validation summaries, review findings, and follow-up links. Local/runner events may enrich replay later, but GitHub remains the durable source of truth.

Requirements

• Reconstruct replay from GitHub data alone for MVP.
• Group events by AgentFlow phases 0-8 and outcome sections: Goal, Scope, Decisions, Validation, Review/Gates, PR Readiness, Outcome, Follow-ups.
• Show evidence links and inferred/missing badges.
• Default to compact summary for short goals; allow drill-down.
• Never expose raw transcripts, prompts, tool inputs, secrets, or full logs.
• Export replay summary as markdown.

Technical Design

Child issues cover:

• Replay: Define Goal Story timeline schema and derivation rules #128 — timeline schema and derivation rules.
• Replay: Build GitHub data adapters for goal reconstruction #129 — GitHub data adapters for goal reconstruction.
• Replay: Implement compact Goal Story UI with filters and evidence drill-down #130 — compact Goal Story UI with filters and evidence drill-down.
• Replay: Add markdown export for Goal Story summaries #131 — markdown export for Goal Story summaries.
• Replay: Validate privacy, accuracy, documentation, and rollout #132 — privacy, accuracy, documentation, and rollout validation.

Acceptance criteria

• Replay epic has linked child issues for implementation, validation, and docs.
• MVP can reconstruct a goal story from GitHub issue/PR data without local runner state.
• Replay labels inferred/missing evidence honestly.
• Replay is read-only and cannot trigger agent execution.
• Privacy requirements explicitly forbid raw transcript/prompt/tool-input display.

Open questions

• Should UI label be “Replay” or “Goal Story” by default?
• Should replay be nested under Cockpit issue view or top-level goal view?
• Which GitHub check/status APIs are required for validation replay in MVP?

Feature Tracking

• Replay: Define Goal Story timeline schema and derivation rules #128 — Timeline schema and derivation rules
• Replay: Build GitHub data adapters for goal reconstruction #129 — GitHub data adapters for goal reconstruction
• Replay: Implement compact Goal Story UI with filters and evidence drill-down #130 — Compact Goal Story UI with filters and evidence drill-down
• Replay: Add markdown export for Goal Story summaries #131 — Markdown export for Goal Story summaries
• Replay: Validate privacy, accuracy, documentation, and rollout #132 — Privacy, accuracy, documentation, and rollout validation

Test plan

• Fixture tests for timeline derivation from GitHub issue/PR/comment data.
• UI tests for compact/expanded replay and filters.
• Privacy tests ensuring forbidden fields are not rendered/exported.
• Export snapshot tests for markdown summaries.

Workflow classification

Standard product feature with privacy-sensitive UI constraints. Requires tester and tech writer phases; security review if remote sharing/export includes private repository data.

[github-actions]

Integrated into development by PR #133: #133

Merge commit: 823d2e784f61e11645d2902e35401134fe0b1bdc
Release/promotion to main is tracked separately.