discovery-review-criteria
Gate definitions and verdict rubric for the DISCOVER reviewer — applied across 3 lenses (completeness, prioritization, duplicate detection) with 6 gates (G1–G6).
When to use
- When the
backlog-discoverer-reviewerruns a review pass on DISCOVER artefacts - To evaluate the quality of triage reports and sprint proposals
- To derive a formal verdict on issue coverage, prioritization coherence, and duplicate handling
Entry contract
triage-report-{date}.md— triage report with classified issues and sprint proposals- Read access to the GitHub repository for the G2 sample-check (open P0/P1 issues)
- Declared team capacity (team-days)
Exit contract
- Formal verdict:
approved,changes_requested, orrejected - Findings per lens with severity (
BLOCKER,HIGH,MEDIUM) - For G2 (BLOCKER): list of P0/P1 issues absent from the triage report
Invariants
- G2 is the only BLOCKER gate — a P0/P1 issue missing from the triage is a critical failure
- Effective capacity = team-days × 0.7 for P1/P2/P3 issues (P0 always overrides capacity)
- XL issues never enter the sprint — G4 fails if an XL issue is included
- No P3 before P1 — a P3 in the sprint while a P1 is excluded is a priority inversion
| Gate | Lens | Severity | Definition |
|---|---|---|---|
| G1 | Completeness | HIGH | All 3 discovery modes applied, or skipped modes justified |
| G2 | Completeness | BLOCKER | No open P0/P1 issue absent from the report (top-5 sample-check) |
| G3 | Prioritization | HIGH | No priority inversions; every P0 has a written justification |
| G4 | Prioritization | HIGH | Effective capacity respected; no XL issue in the sprint |
| G5 | Duplicates | HIGH | No undetected duplicate pair (normalised similarity > 80%) |
| G6 | Duplicates | MEDIUM | All 40–80% pairs flagged in the “Duplicates Detected” section |
Similarity thresholds:
| Level | Similarity | Action required |
|---|---|---|
| EXACT | > 95% | One issue labelled status/duplicate linked to the original |
| NEAR | 80–95% | Merge recommendation; both issues linked and documented |
| RELATED | 40–80% | Flagged as “related”, recommendation documented |
| DIFFERENT | < 40% | No action required |
Why this shape
Discovery is the least visible phase but the most expensive when it fails: a P0 issue absent from the sprint becomes a production incident or critical debt. Gates G2 and G4 protect against the two most frequent failures — invisible critical issues and over-loaded sprints.
« Maximizing the amount of work not done is essential. » — Reinertsen, D. G., The Principles of Product Development Flow, 2009.
Allowed customisation
- Similarity threshold for G5 (L2 — do not lower below 70%)
- Number of issues in the G2 sample-check (L2 — do not lower below 3)
- Severity of G1 and G3 (L2)
See also
- adversarial-review-lenses — Independent-lens verdict procedure
- issue-triage — Triage skill that produces the artefacts reviewed here
- backlog-discoverer — Agent that produces DISCOVER artefacts