| name | trial-approval |
| description | Assess trial Alliance members for promotion to active (full members) using fleet attendance, solo/small-gang kills, and Discord voice, then produce a reviewer report and admin upload CSV. Use when approving trial, moving people off trial, graduating trial to active, reviewing who earned full membership, or running a trial hygiene pass. |
Trial Approval
Promote trial Alliance members to active when they are participating in
the alliance — not fleets alone. Solo / small-gang kills and Discord voice
count alongside fleet attendance. Deliver a report + upload CSV. Do not
change status yourself — the executor uploads after review.
Also read debug-production-readonly-db.
Policy
Trial is provisional. Full membership means they show up as part of the
alliance: alliance fleets, solo / small-gang combat, voice with the community,
or a clear mix.
We do not care just about fleets. We care about them doing things like solo
/ small gang kills, being in voice, etc. Participating in the alliance.
Participation must span the window — do not approve on a strong first month
then silence. Full-window totals alone are not enough; recent activity is
required.
active → affiliation group only (Trial Discord role stripped) via
sync_user_community_groups.
Bias: when unsure, hold trial. Resolve every case — no borderline dump.
Wrong-affiliation trials (Guest / Militia / requires_trial=False) are not
this pass — note them for fix_trial_status_by_affiliation.
This is not on-leave. Leave is fleet-primary (pings /
OPSEC). Trial approval is co-equal participation paths.
Quick start
cd backend
pipenv run python ../.cursor/skills/trial-approval/scripts/fetch_trial_activity.py --json
Defaults: 90d window + 30d recent slice; affiliation Alliance;
status trial.
Workflow
Task Progress:
- [ ] Fetch trial Alliance activity (--json)
- [ ] Split wrong-affiliation (fix command) vs real trial
- [ ] Apply recency gate (30d / days_since_activity)
- [ ] Decide approve / hold; Path + Conf + reason
- [ ] Emit report + CSV (approve rows only)
- [ ] Stop for executor review/upload
Scope
| Filter | Value |
|---|
| Affiliation | Alliance (check prod name if fetch is empty) |
| Status | trial only |
| Activity | All linked EveCharacters (account-level) |
Signals
| Signal | Model | Roll-up |
|---|
| Fleets | EveFleetInstanceMember | Distinct instances across linked EVE IDs |
| Kills | EveCharacterKillmailAttacker | Attackers; exclude self-victim; gang buckets |
| Voice | DiscordChannelActivityRecord | voice_minute by Django username → hours |
Gang buckets (attacker count on the killmail): small ≤10, medium 11–24,
large 25–39, blob 40+. Solo / small-gang = small bucket.
Fetch also emits a 30d recent slice and ages:
| Field | Meaning |
|---|
fleets_30d / kills_30d / kills_small_30d / voice_hours_30d | Same metrics, last 30d only |
days_since_fleet / days_since_kill / days_since_voice | Age of latest event in window (null if none) |
days_since_activity | Min of the three ages (null if fully dark in window) |
Login, mining, PI, industry alone do not clear the bar. Note ESI gaps / zero
linked chars / untracked ops in reasons when they matter.
Decision rules (90d)
Paths are co-equal. Approve when one strong path or two medium paths,
and the recency gate passes.
Strong path (approve, Conf high):
- Fleets ≥ 4
- Small-gang kills ≥ 8 (small bucket)
- Voice ≥ 10h and (fleets ≥ 1 or kills ≥ 3)
Medium path — combine two for approve (Conf medium), or one strong + weak support:
| Path | Medium bar |
|---|
| Fleets | 2–3 |
| Small-gang | 4–7 |
| Total kills (any gang) | ≥ 10 with some small-gang (≥ 3) |
| Voice | ≥ 5h with any fleet or kill touch |
Recency gate (required for approve)
Full-window strength is necessary but not sufficient.
Recent enough to approve when any of:
days_since_activity ≤ 30 (fleet, kill, or voice in the last 30d)- Or explicit 30d touch:
fleets_30d ≥ 1 or kills_30d ≥ 1 or voice_hours_30d ≥ 1
Front-loaded / stale — hold even if 90d path clears:
- 90d totals would approve, but no recent touch (
days_since_activity > 30 or
null with only early-window activity) — they started strong then disappeared
- Typical shape: high
fleets / kills_small over 90d, but fleets_30d = 0,
kills_30d = 0, voice_hours_30d ≈ 0
Reason tag: Front-loaded — … last activity Nd ago (or no activity in 30d).
Conf: do not approve; hold for CEO contact / re-eval when they return.
Do not overweight month-one PAP: if almost all kills/fleets sit in the older
part of the window and the last 30d is quiet, hold.
Hold trial:
- Dark: fleets ≤ 1, kills ≈ 0, voice ≈ 0
- Front-loaded / stale (above)
- Voice-only social ghost: high voice, no fleets, no kills
- Blob-only killboard with no fleets / no small-gang / no voice (unclear alliance play)
- No linked characters (caveat; Conf medium if somehow recommending — prefer hold)
Tag approve Path: Fleet, Small-gang, Voice, or Mixed.
Reason line: Path + 1–2 metrics; add last-activity age when non-obvious. See
examples.md.
Deliverables
Always both. Sort approves by Conf (high then medium).
Summary: N approve (H/M); H held (F front-loaded); W wrong-affiliation. Window: 90d / recent 30d.
Report
| Username | Primary | Previous | New | Fleets | Kills | Small | Voice | 30d | Last | Path | Conf | Reason |
|---|
- Previous =
trial; New = active
- Voice as hours (
12.5h)
- Small = small-gang kill count
- 30d = short recent slice (
2F/5K/1.2h or quiet)
- Last =
days_since_activity (12d / —)
- Held / wrong-affiliation: counts only unless asked; call out front-loaded holds when relevant
CSV (admin upload)
Approve rows only (recency gate already applied):
username,community_status,reason
someuser,active,Mixed — 5 fleets, 9 small-gang, 6h voice (90d; last 8d)
otheruser,active,Small-gang — 0 fleets, 14 small-gang kills (90d; last 3d)
community_status = active. reason ≤255. Username = Django username.
| Flag | Use |
|---|
--days 90 | Lookback |
--affiliation Alliance | AffiliationType.name |
Executor: apply CSV
- Review report (metrics, 30d/Last, Path/Conf/Reason). Treat front-loaded holds as Contact candidates, not silent Passes.
- Delete rows you want to keep on trial; save
trial_approve_YYYY-MM-DD.csv (UTF-8).
- Admin:
/admin/groups/usercommunitystatus/bulk-upload/
- Upload; prefer per-row reasons.
- Upload and apply → Celery; Discord role sync may take minutes.
- Spot-check: status
active, Trial role gone, Alliance role kept.
Bulk upload only. No prod shell writes. Small batches may use admin Approve trial.
Wrong-affiliation list → pipenv run python manage.py fix_trial_status_by_affiliation (with --dry-run first) on an environment that can write — not this skill’s CSV.
Related
