| R01 |
The full 41 GB GitSkills SQLite is attempted instead of the 277 MB sample |
Days lost to downloads and out-of-memory failures; pipeline not reproducible on a laptop |
Ambition, unclear scope decision |
7 |
6 |
3 |
126 |
Sample is the approved population by default; any full-dataset pass needs an ADR (Sprint 1) and runs as a separate streamed step |
Product Owner |
Any PR referencing the full dump |
Closed by ADR-0007 (#68, 2026-10-06): the sample is the population for RQ1 to RQ3; a full-dataset pass is future work that needs a new ADR |
| R02 |
SpecMine sample is non-representative (61 percent of repos have >= 100 stars vs 1.3 percent in the full corpus; 97.5 percent of specs first committed in 2026) |
Findings overgeneralized if SpecMine is used |
Sample was built to be PR-rich, not random |
7 |
9 |
4 |
252 |
P-01 does not use SpecMine; if any SpecMine result is reported, frame it as "in this sample" and state the bias |
Product Owner |
Any SpecMine claim in the report |
Open |
| R03 |
GitHub code search is a lower bound (default branches only, files under 384 KB, active repos, forks only if more starred than parent) |
Population statements are wrong |
Dataset construction limits |
6 |
9 |
4 |
216 |
Define the population as "files indexed by GitHub code search in July 2026"; list the limits in the threats section |
Developers |
Report draft review |
Open |
| R04 |
Commit history in GitSkills follows the current file path, so renames reset first_commit_at and commit_count |
Variant ordering (RQ3) and age metrics biased |
Dataset limitation |
6 |
7 |
5 |
210 |
Document in the data dictionary; treat ordering as uncertain; inspect renamed cases in the validation sample |
Developers |
Any metric using first_commit_at or commit_count |
Open |
| R05 |
A member is unavailable for a week or more (illness, work, family) |
Their sprint items stall; review and validation work waits |
Life; four schedules |
6 |
6 |
3 |
108 |
Every workstream has a lead and a support; stand-ups surface absence early; re-plan at the Thursday check-in; the Thanksgiving break is buffer |
Scrum Master |
Two missed stand-ups in a row by the same member |
Open |
| R06 |
Uneven contribution: one or two members do most of the work |
Individual grades and learning suffer; burnout; rubric evidence of shared work is weak |
Most of the repository was built by one member before the group was reinstated; unclear ownership |
7 |
7 |
3 |
147 |
Work sign-up rules (docs/workspace/WORK_SIGNUP.md); contribution-balance KPI with band 0.15 to 0.45 read at every check-in; escalation path in TEAM_CHARTER.md section 10 |
Scrum Master |
Any member's contribution share outside the band at a check-in |
Open |
| R07 |
Scope creep after topic selection (a second question, a second dataset, a model not needed) |
Nothing finishes; report is shallow |
Enthusiasm, no Definition of Ready |
8 |
7 |
3 |
168 |
New scope must be an issue approved at planning; added_after_planning reported per sprint |
Product Owner |
added_after_planning above 20 percent of committed items |
Open |
| R08 |
Pipeline does not run on a clean machine (missing dependency, hard-coded path, undocumented download step) |
Sprint 3 and final acceptance criteria fail |
Code is only run in its author's environment |
9 |
7 |
4 |
252 |
CI runs tests from scratch; gemba walk at every sprint review from a fresh clone, by a member who did not write the README |
Developers |
CI failure or gemba walk failure |
Open |
| R09 |
A member executes scripts or commands bundled in dataset artifacts |
Security incident, academic integrity issue; rubric explicitly forbids it |
Curiosity, copy-paste while debugging |
10 |
3 |
8 |
240 |
Dataset content is data, never code; loaders read content as text only; static analysis only; no subprocess on dataset content in src/ (checked in review) |
Scrum Master |
Any PR that opens dataset content with anything other than a text reader |
Open |
| R10 |
AI-use log incomplete or disclosure missing from the report |
Responsible AI criterion failed |
Forgetting to log; unclear what counts |
8 |
6 |
6 |
288 |
ai-use-log.md entry is a PR template checkbox and a reviewer checklist item; checked at each Thursday check-in; disclosure paragraph is a report section from Sprint 2 |
Scrum Master |
PR merged with the checkbox unticked |
Open |
| R11 |
Rater disagreement or loss of rater independence (low inter-rater kappa, or the second rater sees the primary rater's labels) |
Validation cannot support the result; precision overstated or unreliable |
Vague definitions; new raters with little practice; primary label files visible in the repository |
7 |
5 |
4 |
140 |
Written guideline and practice items before labelling; blindness rule in docs/validation/README.md; second-rater PR contains only their own label file; adjudication in error analysis; kappa reported |
Product Owner |
Inter-rater kappa below 0.6, or a primary-rater label file changed before the second rater's file for that kind is committed |
Open |
| R12 |
Reviews are delayed, so PRs wait on the one required approval |
Work piles up unmerged; merge conflicts; sprint items finish late |
Four schedules; reviews treated as lower priority than own work |
5 |
6 |
3 |
90 |
48-hour review SLA; reviewer rotation; first-review KPI read at every check-in; Friday week summary lists PRs waiting more than 48 hours |
Scrum Master |
Median first-review turnaround above 48 hours for two weeks |
Open |
| R13 |
OneDrive sync conflicts with the git working tree (locked files, "conflicted copy" files, corrupted .git) |
Lost commits, confusing diffs |
The repository folder lives inside a synced OneDrive directory |
6 |
7 |
3 |
126 |
Clone the repository outside OneDrive for daily work (for example C:\dev\); never commit files named *-conflicted copy* |
Developers |
Any file with "conflicted copy" in the name |
Open |
| R14 |
Thanksgiving gap (Nov 19 to Nov 29) between Sprint 3 and Finalization loses momentum |
Finalization week is a scramble |
Calendar |
6 |
8 |
2 |
96 |
Sprint 3 review produces a finalization checklist with owners and dates before the break; one stand-up on Nov 25 |
Scrum Master |
Sprint 3 review without a finalization checklist |
Open |
| R15 |
Instructor adjusts sprint dates |
Milestones, timeline, and metrics windows are wrong |
Institutional calendar |
4 |
5 |
1 |
20 |
Dates live in one place (project.yml); a change updates milestones via scripts/bootstrap_github.py and is recorded in the decision log |
Scrum Master |
Any date change announced in class |
Open |
| R16 |
Members onboard late in formation (invited 2026-09-15, one day before the kickoff) |
Charter signatures, setup, and first contributions slip into Sprint 1; Sprint 1 capacity is overestimated |
Group reinstated late in formation week (ADR-0006) |
6 |
6 |
3 |
108 |
Onboarding path in docs/GETTING_STARTED.md; onboarding issue form; pairing during setup; Sprint 1 item S1-12; first week sized for ramp-up |
Scrum Master |
Any onboarding issue still open at Sprint 1 planning on 2026-09-17 |
Open |