feat: add /codeboarding-feedback command to capture PR feedback via PostHog

[brovatten]

What

Adds a lightweight /codeboarding-feedback <message> PR-comment command. When a comment's first word is /codeboarding-feedback, the action sends the message text and PR context to CodeBoarding via PostHog as an explicit codeboarding_feedback_submitted event, reacts 👍 to the comment, and exits before any checkout, engine setup, cache restore, or LLM-key handling — it never runs the expensive analysis path.

Implements feedback_plan.md.

How

• scripts/submit_feedback.py — stdlib-only (no new deps; runs before engine setup). Strips the command token (internal newlines preserved, outer whitespace trimmed), caps text at feedback_max_chars, POSTs one event to PostHog /i/v0/e/, swallows every send failure, and never prints the feedback body. Mirrors Core's telemetry/service.py: same default public ingest key, CODEBOARDING_POSTHOG_KEY/CODEBOARDING_POSTHOG_HOST overrides, and DO_NOT_TRACK / CODEBOARDING_TELEMETRY=false opt-outs.
• action.yml — new feedback_command (default /codeboarding-feedback) and feedback_max_chars (default 4000) inputs; an early feedback branch in the guard step (before the trigger/trusted-collaborator checks) that runs the script and exits with skip=true + feedback_received=true; a new Acknowledge feedback step gated on feedback_received. The existing Acknowledge command step stays scoped to analysis (skip != 'true').
• README.md — a "Feedback command" subsection, two Inputs-table rows, and opt-out docs that state plainly this is explicit user-submitted feedback (the text you wrote + who wrote it + PR context), not anonymous telemetry.
• tests/test_submit_feedback.py — 26 stdlib unittest cases.

Event

codeboarding_feedback_submitted, distinct_id = github-user:<sender_id> (fallback github-run:<run_id>), source = github_action_feedback, with feedback_text (capped), feedback_length, feedback_truncated, repository, repository_id, pr_number, comment_id, comment_url, author_association, sender_login, sender_id, run_id, run_attempt, action_ref.

Trust model

The feedback branch runs no PR-head code and sends no source, so — per the plan — it is intentionally not gated on author association; the calling workflow's own if: condition is the control point. The dogfood workflow already restricts comment triggers to OWNER/MEMBER/COLLABORATOR, and its startsWith('/codeboarding') filter already matches /codeboarding-feedback, so no workflow change is needed.

Verification

• 149/149 unit tests pass; black clean (120 cols); action.yml YAML parses; shellcheck clean; actionlint adds zero findings.
• End-to-end subprocess test of the real entry point: opt-out → no send; empty body → nothing sent; unreachable host → graceful swallow, exit 0, no feedback-text leak.

Notes

• Privacy: unlike Core telemetry, this event deliberately includes repository, sender_login, and the user-written feedback_text. That is appropriate for explicit feedback and is documented in the README.
• V1 posts directly to PostHog (public ingest key). A future CodeBoarding-owned intake endpoint could add rate-limiting, redaction, and routing to Slack/Discord/Linear/Issues.

[codeboarding-review]

Architecture review · 14 components changed

graph LR
    n_Analysis_Orchestrator["Analysis Orchestrator"]
    n_Structural_Diffing_Engine["Structural Diffing Engine"]
    n_Mermaid_Visualization_Engine["Mermaid Visualization Engine"]
    n_UX_Integration_Layer["UX #amp; Integration Layer"]
    n_Analysis_Orchestrator -- "orchestrates codebase scanning and diff extract…" --> n_Structural_Diffing_Engine
    n_Analysis_Orchestrator -- "provides analysis metadata for final reporting" --> n_UX_Integration_Layer
    n_Structural_Diffing_Engine -- "provides structural change data for diagram gen…" --> n_Mermaid_Visualization_Engine
    n_Mermaid_Visualization_Engine -- "queries specific file changes for component ren…" --> n_Structural_Diffing_Engine
    n_Mermaid_Visualization_Engine -- "provides rendered diagrams for GitHub comments" --> n_UX_Integration_Layer
    n_UX_Integration_Layer -- "triggers feedback loops and telemetry based on…" --> n_Analysis_Orchestrator
    classDef added fill:#1f883d,stroke:#0b5d23,color:#ffffff;
    classDef modified fill:#bf8700,stroke:#7d4e00,color:#ffffff;
    classDef deleted fill:#cf222e,stroke:#82071e,color:#ffffff,stroke-dasharray:5 3;
    class n_Structural_Diffing_Engine,n_UX_Integration_Layer modified;
    linkStyle 5 stroke:#0b5d23,stroke-width:2px;
    linkStyle 0,1,2,3,4 stroke:#7d4e00,stroke-width:2px;

Loading

Colors indicate component changes compared to main: 🟩 Added · 🟨 Modified · 🟥 Removed

Structural Diffing Engine : 1 file changed

• scripts/diff_to_mermaid.py

UX & Integration Layer : 1 file changed

• scripts/submit_feedback.py

---

Explore this PR’s architecture in your browser or VS Code.

_{codeboarding-action · run 27483242044}