Sentinel — Validation Plan

Validation plan — Sentinel by Aftergraph

Wedge guardrail: Sentinel turns pull requests into merge-ready verdicts. First wedge: CLI review on exact HEAD → SHIP / DO NOT SHIP verdict with cited evidence → GitHub App.

Method: desk research + prototype validation (no outreach pipeline)

Competitor desk research. Confirm all [VERIFY] markers in competitor-analysis.md against first-party vendor docs (CodeRabbit, Greptile, Copilot, Sonar, CodeScene). One afternoon, produces the publishable comparison.
Rule-pack gap analysis. List 20 merge-blocking rules from public postmortems (unauth routes, missing rollbacks, secret leaks, N+1, unindexed migrations). If Sentinel’s v0 rule model covers 15+, the wedge holds.
Prototype validation (5 real PRs). Run the verdict-card mock + manual CLI-shape review against 5 real Aftergraph PRs (1 clean, 2 with real findings, 1 rebased mid-review, 1 dependency bump). Score: verdict correct? evidence cited file:line? rebase invalidated? Record misses publicly in the repo.
Passive waitlist. Landing page with verdict-card visual + “CLI first, App second” + email capture. No launch, no ads — measures pull.
Legal/compliance desk check. What makes a verdict record auditor-acceptable (immutability, export format, retention)? One memo, informs data-model-v0, not a feature yet.

Interviews / pilot: OPTIONAL LATER

Direct interviews, outreach copy, pilot pipeline are explicitly not next actions — optional later. Reactivate only after prototype validation scores 4/5 or better and the waitlist shows pull.

Source: Aftergraph/sentinel@beececfb4885d8055cc6694e8979b47759a212a4 docs/validation-plan.md.md — verbatim, pinned, no reinterpretation.

Was this page helpful?