@wonsukchoi/crondex 0.21.0 → 0.22.0

package/jobs/healthcare/clinical-supply-low-stock-check.yaml

@@ -0,0 +1,30 @@
+id: clinical-supply-low-stock-check
+version: 1
+name: Clinical Supply and PPE Low-Stock Check
+description: >
+  Scans clinical supply and PPE inventory (gloves, masks, gowns, syringes,
+  exam table paper, sanitizer) against reorder thresholds and flags what's
+  running low. Use this if a clinic has ever discovered it was out of
+  gloves or exam paper mid-morning because nobody was watching stock levels
+  between deliveries.
+category: healthcare
+tags: [healthcare, inventory, supplies, ppe, operations]
+schedule: "0 7 * * 1,4"
+timezone: "America/New_York"
+runner: shell
+command: >
+  awk -F',' '
+    NR==1 {next}
+    ($3 + 0) <= ($4 + 0) {
+      printf "LOW STOCK: %s (%s) — %d on hand, reorder threshold %d\n", $1, $2, $3, $4;
+    }
+  ' "{{supply_inventory_csv_path}}"
+variables:
+  supply_inventory_csv_path:
+    default: "./clinical-supply-inventory.csv"
+    description: "Path to a CSV with columns: item_name,category,current_qty,reorder_threshold."
+compatible_agents: [generic, claude, codex]
+notes: >
+  Zero-token, read-only check. Assumes a maintained inventory export (manual count
+  or supply-management software) — this is the "did anyone check the shelf" backstop,
+  it doesn't place the reorder itself.

package/jobs/healthcare/controlled-substance-inventory-reconciliation.yaml

@@ -0,0 +1,42 @@
+id: controlled-substance-inventory-reconciliation
+version: 1
+name: Controlled Substance Inventory Reconciliation
+description: >
+  Reviews the practice's controlled-substance dispensing/administration log
+  (e.g. in-office injectable opioids, ketamine, testosterone, Schedule II-V
+  medications kept on site) for gaps or unreconciled counts. Use this if a
+  discrepancy in a controlled-drug count has ever gone unnoticed for weeks
+  in a human practice instead of being caught same-day — this covers a
+  clinic's own in-office stock, the same DEA-style pattern as
+  jobs/veterinary/controlled-substance-log-audit.yaml but for a human
+  medical/dental practice rather than an animal clinic, so both are kept as
+  separate jobs.
+category: healthcare
+tags: [healthcare, controlled-substances, compliance, audit, dea]
+schedule: "0 20 * * *"
+timezone: "America/New_York"
+runner: agent-prompt
+prompt: |
+  Read today's controlled-substance log from {{log_source}} for
+  {{substances_tracked}}.
+  1. Check that every entry has a matching patient/encounter reference —
+     flag any entry missing one.
+  2. Compare the running balance against the last physical count on record;
+     flag any mismatch.
+  3. Flag any day with zero entries logged if the practice was open and
+     controlled substances are known to be in regular use — a suspiciously
+     quiet log is itself worth a second look.
+  4. If everything reconciles, say so in one line.
+variables:
+  log_source:
+    default: "your controlled-substance log (paper log photographed/transcribed, or e-log system)"
+    description: Where the controlled-substance usage log is recorded.
+  substances_tracked:
+    default: "injectable opioids, ketamine, and other Schedule II-V substances kept on site"
+    description: Which substances this log covers.
+compatible_agents: [claude, codex, hermes, openclaw, generic]
+notes: >
+  Read-only reconciliation check — does not replace your practice's actual
+  DEA-compliant logging system, and any discrepancy found should go through your
+  normal reporting/investigation process. Assumes an already-exported or
+  transcribed log rather than live PHI system access.

package/jobs/healthcare/hipaa-access-log-audit.yaml

@@ -0,0 +1,74 @@
+id: hipaa-access-log-audit
+version: 1
+name: HIPAA Chart Access Log Audit
+description: >
+  Scans the EHR's chart-access log for high-volume, off-hours, or otherwise
+  unusual access patterns and has an agent judge which are worth a closer
+  look. Use this if unauthorized or curious chart access (a coworker's
+  family member, a local celebrity, an ex-spouse) has ever gone unnoticed
+  because nobody was reviewing the access log until an audit or complaint
+  forced it.
+category: healthcare
+tags: [healthcare, hipaa, compliance, audit, security]
+schedule: "0 7 * * 1"
+timezone: "America/New_York"
+runner: hybrid
+command: >
+  awk -F',' -v threshold="{{daily_access_threshold}}" -v off_start="{{off_hours_start}}" -v off_end="{{off_hours_end}}" '
+    NR==1 {next}
+    {
+      day = substr($3,1,10); hour = substr($3,12,2) + 0;
+      key = $1 SUBSEP day;
+      count[key]++;
+      staffname[key] = $1; dayval[key] = day;
+      if (hour >= off_start || hour < off_end) {
+        printf "OFF-HOURS ACCESS: %s accessed %s at %s (action: %s)\n", $1, $2, $3, $4;
+      }
+    }
+    END {
+      for (k in count) {
+        if (count[k] > threshold) {
+          printf "HIGH-VOLUME ACCESS: %s accessed %d distinct log entries on %s (threshold %d)\n", staffname[k], count[k], dayval[k], threshold;
+        }
+      }
+    }
+  ' "{{access_log_csv_path}}"
+prompt: |
+  Run the flagged entries from the access-log scan above (high-volume days
+  and off-hours access events) through a closer read.
+  1. For each flagged staff member/day, look at which patient records were
+     touched and whether the access pattern looks consistent with normal
+     clinical or front-desk duties for that role (e.g. a nurse pulling
+     charts for today's patient panel) versus inconsistent with it (e.g.
+     access to a patient not on that staff member's schedule, repeated
+     access to the same non-assigned patient, or a name match suggesting a
+     coworker, family member, or public figure).
+  2. Rank flags from "likely benign, no action" to "recommend compliance
+     officer review."
+  3. Do not name a possible relationship as fact — describe the pattern
+     observed and let a human confirm it.
+  4. If nothing stands out beyond routine volume, say so in one line.
+variables:
+  access_log_csv_path:
+    default: "./chart-access-log.csv"
+    description: "Path to a CSV with columns: staff_name,patient_name,access_timestamp (YYYY-MM-DD HH:MM),action (view/edit/print)."
+  daily_access_threshold:
+    default: 40
+    description: Number of chart access events by one staff member in one day that triggers a high-volume flag.
+  off_hours_start:
+    default: 20
+    description: Hour (0-23, local clinic time) after which access counts as off-hours.
+  off_hours_end:
+    default: 6
+    description: Hour (0-23) before which access still counts as off-hours.
+compatible_agents: [claude, codex, hermes, openclaw, generic]
+script_note: >
+  The command gives you the raw high-volume and off-hours flags for free, purely
+  from counts and timestamps. Only the prompt pass can judge whether a flagged
+  pattern looks like inappropriate snooping versus legitimate clinical coverage —
+  that judgment call is what the LLM pass adds.
+notes: >
+  Assumes an already-exported, de-identified-where-possible access log rather than
+  live EHR/PHI system access — export staff name, patient identifier, timestamp,
+  and action only. Any confirmed concern should go through your practice's actual
+  HIPAA compliance/privacy officer process, not this job.

package/jobs/healthcare/patient-balance-collections-aging-check.yaml

@@ -0,0 +1,41 @@
+id: patient-balance-collections-aging-check
+version: 1
+name: Overdue Patient Balance Aging Check
+description: >
+  Buckets outstanding self-pay patient balances by how long they've been
+  unpaid (30/60/90+ days) and flags the oldest ones for collections
+  follow-up. Use this if patient balances have ever quietly aged past the
+  point of easy recovery because nobody was reviewing accounts receivable
+  by age, only by total dollar amount.
+category: healthcare
+tags: [healthcare, billing, collections, accounts-receivable, patients]
+schedule: "0 8 1,15 * *"
+timezone: "America/New_York"
+runner: shell
+command: >
+  today=$(date +%s);
+  awk -F',' -v today="$today" -v flag_days="{{flag_after_days}}" '
+    NR==1 {next}
+    {
+      cmd = "date -d \"" $3 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $3 "\" +%s 2>/dev/null";
+      cmd | getline invoice_epoch; close(cmd);
+      age_days = (today - invoice_epoch) / 86400;
+      bucket = (age_days < 30) ? "current" : (age_days < 60) ? "30-59" : (age_days < 90) ? "60-89" : "90+";
+      if (age_days >= flag_days) {
+        printf "OVERDUE BALANCE: %s — $%s, invoiced %s (%d days, bucket %s)\n", $1, $2, $3, age_days, bucket;
+      }
+    }
+  ' "{{patient_balances_csv_path}}"
+variables:
+  patient_balances_csv_path:
+    default: "./patient-balances.csv"
+    description: "Path to a CSV with columns: patient_name,balance_amount,invoice_date (YYYY-MM-DD)."
+  flag_after_days:
+    default: 60
+    description: Days since invoicing before an unpaid balance is flagged for collections follow-up.
+compatible_agents: [generic, claude, codex]
+notes: >
+  Assumes an already-exported CSV from your billing/practice-management system
+  rather than live PHI system access. Read-only aging report — actual collections
+  outreach should follow your practice's billing/compliance policy (e.g. FDCPA-style
+  notice requirements where applicable).

package/jobs/healthcare/patient-immunization-schedule-check.yaml

@@ -0,0 +1,41 @@
+id: patient-immunization-schedule-check
+version: 1
+name: Patient Immunization Schedule Check
+description: >
+  Checks which patients are due or overdue for a scheduled immunization and
+  lists them for outreach. Use this if a patient (child or adult) has ever
+  fallen behind on a vaccine series because nobody was tracking due dates
+  against the immunization schedule between visits — this is for human
+  patients, distinct from jobs/veterinary/pet-vaccine-due-reminder.yaml
+  which tracks pet vaccines at an animal clinic.
+category: healthcare
+tags: [healthcare, immunization, vaccine, patients, preventive-care]
+schedule: "0 8 * * *"
+timezone: "America/New_York"
+runner: shell
+command: >
+  awk -F',' -v days="{{warn_days_before}}" '
+    NR==1 {next}
+    {
+      cmd = "date -d \"" $3 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $3 "\" +%s 2>/dev/null";
+      cmd | getline due_epoch; close(cmd);
+      "date +%s" | getline now_epoch; close("date +%s");
+      diff_days = (due_epoch - now_epoch) / 86400;
+      if (diff_days <= days) {
+        printf "IMMUNIZATION DUE: %s — %s due %s (%d days)\n", $1, $4, $3, diff_days;
+      }
+    }
+  ' "{{patients_csv_path}}"
+variables:
+  patients_csv_path:
+    default: "./patient-immunizations.csv"
+    description: "Path to a CSV with columns: patient_name,date_of_birth,due_date (YYYY-MM-DD),vaccine_name."
+  warn_days_before:
+    default: 21
+    description: How many days before (or past) the due date to start listing.
+compatible_agents: [generic, claude, codex]
+notes: >
+  Assumes an already-exported CSV from your EHR/immunization registry rather than
+  live PHI system access. Read-only backstop check — cross-verify against your
+  state immunization registry before contacting patients, since due dates can shift
+  based on catch-up schedules.

package/jobs/healthcare/patient-portal-message-backlog-check.yaml

@@ -0,0 +1,39 @@
+id: patient-portal-message-backlog-check
+version: 1
+name: Patient Portal Message Backlog Check
+description: >
+  Checks unanswered patient portal messages against a response-time SLA and
+  flags anything overdue. Use this if a patient message has ever sat
+  unanswered in the portal inbox for days because it scrolled off the top
+  of a busy staff member's queue and nobody was tracking age.
+category: healthcare
+tags: [healthcare, patient-portal, messaging, sla, patients]
+schedule: "0 */2 * * 1-6"
+timezone: "America/New_York"
+runner: shell
+command: >
+  now=$(date +%s);
+  awk -F',' -v now="$now" -v sla="{{sla_hours}}" '
+    NR==1 {next}
+    $4 == "unanswered" {
+      cmd = "date -d \"" $2 "\" +%s 2>/dev/null || date -j -f \"%Y-%m-%d %H:%M\" \"" $2 "\" +%s 2>/dev/null";
+      cmd | getline received_epoch; close(cmd);
+      age_hours = (now - received_epoch) / 3600;
+      if (age_hours >= sla) {
+        printf "OVERDUE MESSAGE: %s — \"%s\" received %s (%d hrs ago, unanswered)\n", $1, $3, $2, age_hours;
+      }
+    }
+  ' "{{portal_messages_csv_path}}"
+variables:
+  portal_messages_csv_path:
+    default: "./portal-messages.csv"
+    description: "Path to a CSV with columns: patient_name,received_at (YYYY-MM-DD HH:MM),subject,status (unanswered/answered)."
+  sla_hours:
+    default: 24
+    description: Hours a message can go unanswered before it's flagged overdue.
+compatible_agents: [generic, claude, codex]
+notes: >
+  Assumes an already-exported CSV from your patient portal/EHR messaging module
+  rather than live access to message content — export subject lines only if full
+  message bodies raise PHI-export concerns for your practice. Read-only — does not
+  reply to patients.

package/jobs/healthcare/patient-satisfaction-survey-tracking.yaml

@@ -0,0 +1,59 @@
+id: patient-satisfaction-survey-tracking
+version: 1
+name: Patient Satisfaction Survey Response Tracking
+description: >
+  Tallies patient satisfaction survey response rates and surfaces low
+  scores with comments for follow-up. Use this if low-scoring survey
+  responses have ever sat unread in an inbox, or a slipping response rate
+  went unnoticed for months — distinct from generic online reviews, this
+  tracks a practice's own post-visit survey instrument.
+category: healthcare
+tags: [healthcare, patient-satisfaction, survey, feedback, patients]
+schedule: "0 9 * * 1"
+timezone: "America/New_York"
+runner: hybrid
+command: >
+  awk -F',' -v low="{{low_score_threshold}}" '
+    NR==1 {next}
+    { sent++; }
+    $3 != "" { responded++; }
+    ($3 != "" && $4 != "" && ($4+0) <= low) {
+      printf "LOW SCORE: %s — score %s: \"%s\" (responded %s)\n", $1, $4, $5, $3;
+    }
+    END {
+      rate = (sent > 0) ? (responded / sent * 100) : 0;
+      printf "RESPONSE RATE: %d/%d sent (%.1f%%)\n", responded, sent, rate;
+    }
+  ' "{{survey_csv_path}}"
+prompt: |
+  Look at the low-score rows and comments flagged by the tally above (from
+  {{survey_csv_path}}).
+  1. Group the comments into common themes (e.g. wait time, billing
+     confusion, communication, front-desk experience).
+  2. For each theme, note how many respondents mentioned it and give one
+     representative quote.
+  3. For any response that names a specific, addressable problem, draft a
+     short, warm outreach note to that patient acknowledging the feedback —
+     draft-only, for staff to review before sending.
+  4. If the response rate this run is notably lower than
+     {{expected_response_rate}}%, flag that as worth investigating
+     separately from the score themes.
+variables:
+  survey_csv_path:
+    default: "./patient-satisfaction-surveys.csv"
+    description: "Path to a CSV with columns: patient_name,visit_date,response_date (blank if no response),score (1-5, blank if no response),comment."
+  low_score_threshold:
+    default: 3
+    description: Scores at or below this value are flagged as low.
+  expected_response_rate:
+    default: 25
+    description: Expected survey response rate (%) — used only as a reference point for flagging an unusually low run.
+compatible_agents: [claude, codex, hermes, openclaw, generic]
+script_note: >
+  The command gives you the response rate and raw low-score rows for free. Only the
+  prompt pass groups comments into themes and drafts a tactful outreach note to an
+  unhappy patient — that synthesis is what the LLM pass adds.
+notes: >
+  Assumes an already-exported CSV from your survey tool (e.g. post-visit SMS/email
+  survey) rather than live PHI system access. Outreach drafts must be reviewed by
+  staff before sending.

package/jobs/healthcare/prescription-refill-request-backlog-check.yaml

@@ -0,0 +1,38 @@
+id: prescription-refill-request-backlog-check
+version: 1
+name: Prescription Refill Request Backlog Check
+description: >
+  Checks pending prescription refill requests against how long they've been
+  waiting and flags anything past a reasonable turnaround. Use this if a
+  refill request has ever sat in a queue long enough that a patient ran out
+  of a maintenance medication before the practice got to it.
+category: healthcare
+tags: [healthcare, prescriptions, refills, patients, medication]
+schedule: "0 9,15 * * 1-5"
+timezone: "America/New_York"
+runner: shell
+command: >
+  today=$(date +%s);
+  awk -F',' -v today="$today" -v hours="{{escalate_after_hours}}" '
+    NR==1 {next}
+    $4 == "pending" {
+      cmd = "date -d \"" $3 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $3 "\" +%s 2>/dev/null";
+      cmd | getline requested_epoch; close(cmd);
+      age_hours = (today - requested_epoch) / 3600;
+      if (age_hours >= hours) {
+        printf "REFILL PENDING: %s — %s requested %s (%d hrs ago)\n", $1, $2, $3, age_hours;
+      }
+    }
+  ' "{{refill_requests_csv_path}}"
+variables:
+  refill_requests_csv_path:
+    default: "./refill-requests.csv"
+    description: "Path to a CSV with columns: patient_name,medication,requested_date (YYYY-MM-DD),status (pending/approved/denied)."
+  escalate_after_hours:
+    default: 48
+    description: Hours a refill request can sit pending before it's flagged.
+compatible_agents: [generic, claude, codex]
+notes: >
+  Assumes an already-exported CSV from your e-prescribing/EHR system rather than
+  live PHI system access. Read-only triage — actual approval/denial of a refill is a
+  clinical decision made by the prescriber, not this job.

package/jobs/healthcare/prior-authorization-expiry-check.yaml

@@ -0,0 +1,53 @@
+id: prior-authorization-expiry-check
+version: 1
+name: Prior Authorization Expiry Check
+description: >
+  Checks treatment-specific prior-authorization requests for approvals about
+  to expire before the procedure happens, and for requests still stuck
+  awaiting payer response. Use this if a procedure has ever been delayed or
+  denied because the prior auth expired, or the payer never responded and
+  nobody was tracking it — distinct from insurance-eligibility-verification-check,
+  which covers general visit eligibility rather than a specific treatment's
+  authorization.
+category: healthcare
+tags: [healthcare, prior-authorization, billing, insurance, compliance]
+schedule: "0 9 * * 1-5"
+timezone: "America/New_York"
+runner: shell
+command: >
+  today=$(date +%s);
+  awk -F',' -v today="$today" -v warn_days="{{warn_days_before_expiry}}" -v stale_days="{{pending_stale_days}}" '
+    NR==1 {next}
+    {
+      if ($4 == "approved" && $6 != "") {
+        cmd = "date -d \"" $6 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $6 "\" +%s 2>/dev/null";
+        cmd | getline expiry_epoch; close(cmd);
+        diff_days = (expiry_epoch - today) / 86400;
+        if (diff_days <= warn_days) {
+          printf "AUTH EXPIRING: %s — %s (%s) expires %s (%d days)\n", $1, $2, $3, $6, diff_days;
+        }
+      } else if ($4 == "pending") {
+        cmd = "date -d \"" $5 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $5 "\" +%s 2>/dev/null";
+        cmd | getline submitted_epoch; close(cmd);
+        pending_days = (today - submitted_epoch) / 86400;
+        if (pending_days >= stale_days) {
+          printf "AUTH STUCK PENDING: %s — %s (%s) submitted %s (%d days, no payer decision)\n", $1, $2, $3, $5, pending_days;
+        }
+      }
+    }
+  ' "{{prior_auth_csv_path}}"
+variables:
+  prior_auth_csv_path:
+    default: "./prior-authorizations.csv"
+    description: "Path to a CSV with columns: patient_name,procedure,payer,status (pending/approved/denied),submitted_date (YYYY-MM-DD),auth_expiry_date (YYYY-MM-DD, blank if still pending)."
+  warn_days_before_expiry:
+    default: 14
+    description: How many days before an approved auth expires to start warning.
+  pending_stale_days:
+    default: 10
+    description: Days a request can sit pending with no payer decision before it's flagged as stuck.
+compatible_agents: [generic, claude, codex]
+notes: >
+  Assumes an already-exported CSV from your billing/prior-auth tracking system rather
+  than live EHR or payer-portal access. Read-only — this flags what needs a call to
+  the payer, it doesn't submit or chase the request itself.

package/jobs/healthcare/staff-cme-credit-tracking.yaml

@@ -0,0 +1,41 @@
+id: staff-cme-credit-tracking
+version: 1
+name: Staff CME Credit Tracking
+description: >
+  Checks each clinician's continuing medical education (CME) credit hours
+  completed against what's required by their cycle end date and warns
+  before anyone falls short. Use this if a provider has ever discovered
+  close to renewal that they were short on CME hours needed to keep their
+  license active — distinct from license-credential-expiry-check, which
+  tracks the license's own expiry date rather than the ongoing education
+  requirement to renew it.
+category: healthcare
+tags: [healthcare, cme, compliance, licensing, staff]
+schedule: "0 8 1 * *"
+timezone: "America/New_York"
+runner: shell
+command: >
+  today=$(date +%s);
+  awk -F',' -v today="$today" -v warn_days="{{warn_days_before_cycle_end}}" '
+    NR==1 {next}
+    {
+      cmd = "date -d \"" $5 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $5 "\" +%s 2>/dev/null";
+      cmd | getline cycle_end_epoch; close(cmd);
+      diff_days = (cycle_end_epoch - today) / 86400;
+      shortfall = ($4 + 0) - ($3 + 0);
+      if (diff_days <= warn_days && shortfall > 0) {
+        printf "CME SHORTFALL: %s — %d/%d hours completed, %d short, cycle ends %s (%d days)\n", $1, $3, $4, shortfall, $5, diff_days;
+      }
+    }
+  ' "{{cme_tracking_csv_path}}"
+variables:
+  cme_tracking_csv_path:
+    default: "./staff-cme-credits.csv"
+    description: "Path to a CSV with columns: staff_name,license_type,cme_hours_completed,cme_hours_required,cycle_end_date (YYYY-MM-DD)."
+  warn_days_before_cycle_end:
+    default: 90
+    description: How many days before the CME cycle ends to start warning about a shortfall.
+compatible_agents: [generic, claude, codex]
+notes: >
+  Assumes a maintained CSV export from your CME tracking or HR/credentialing
+  system, updated whenever a provider logs completed CME hours.

package/jobs/healthcare/telehealth-session-issue-check.yaml

@@ -0,0 +1,51 @@
+id: telehealth-session-issue-check
+version: 1
+name: Telehealth Session No-Show/Tech-Failure Check
+description: >
+  Reviews recent telehealth appointments for no-shows and technical-failure
+  disconnects, and flags patients hitting either problem repeatedly. Use
+  this if a telehealth visit has ever quietly failed to happen — patient
+  never joined, or the call dropped and was never rebooked — with nobody
+  tracking it the way an in-person no-show would normally get tracked.
+category: healthcare
+tags: [healthcare, telehealth, no-show, appointments, patients]
+schedule: "0 18 * * *"
+timezone: "America/New_York"
+runner: shell
+command: >
+  today=$(date +%s);
+  awk -F',' -v today="$today" -v lookback="{{lookback_days}}" -v repeat_threshold="{{repeat_issue_threshold}}" '
+    NR==1 {next}
+    {
+      cmd = "date -d \"" $2 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $2 "\" +%s 2>/dev/null";
+      cmd | getline appt_epoch; close(cmd);
+      age_days = (today - appt_epoch) / 86400;
+      if (age_days < 0 || age_days > lookback) next;
+      if ($3 == "no-show" || $3 == "tech-failure" || $3 == "disconnected") {
+        printf "SESSION ISSUE: %s — %s on %s (%s)\n", $1, $3, $2, $4;
+        count[$1]++;
+      }
+    }
+    END {
+      for (p in count) {
+        if (count[p] >= repeat_threshold) {
+          printf "REPEAT ISSUE: %s has had %d telehealth issues in the last %d days\n", p, count[p], lookback;
+        }
+      }
+    }
+  ' "{{telehealth_sessions_csv_path}}"
+variables:
+  telehealth_sessions_csv_path:
+    default: "./telehealth-sessions.csv"
+    description: "Path to a CSV with columns: patient_name,appointment_date (YYYY-MM-DD),status (completed/no-show/tech-failure/disconnected),provider_name."
+  lookback_days:
+    default: 7
+    description: How many days back to review telehealth sessions.
+  repeat_issue_threshold:
+    default: 2
+    description: Number of session issues within the lookback window before a patient is flagged as a repeat case (may need an in-person alternative offered).
+compatible_agents: [generic, claude, codex]
+notes: >
+  Assumes an already-exported CSV from your telehealth platform or scheduling
+  system rather than live PHI system access. Read-only — repeat cases are worth
+  offering an in-person or phone alternative rather than retrying video indefinitely.

package/jobs/hr/aca-fte-threshold-monitor.yaml

@@ -0,0 +1,44 @@
+id: aca-fte-threshold-monitor
+version: 1
+name: ACA Full-Time-Equivalent Threshold Monitor
+description: >
+  Estimates the company's total ACA full-time-equivalent count from a
+  headcount/hours export and warns when it's approaching the Applicable
+  Large Employer threshold. Use this if the company has ever crossed 50
+  FTEs without anyone noticing until a compliance filing requirement
+  suddenly applied.
+category: hr
+tags: [hr, aca, compliance, headcount]
+schedule: "0 8 1 * *"
+timezone: "America/New_York"
+runner: shell
+command: >
+  awk -F',' -v threshold="{{fte_threshold}}" -v buffer="{{warn_buffer}}" '
+    NR==1 {next}
+    $2 == "full-time" {ft_count++; next}
+    {pt_hours += $3}
+    END {
+      pt_fte = pt_hours / 120;
+      total_fte = ft_count + pt_fte;
+      printf "Full-time headcount: %d, part-time/variable FTE-equivalent: %.1f, total ACA FTE estimate: %.1f\n", ft_count, pt_fte, total_fte;
+      if (total_fte >= (threshold - buffer)) {
+        printf "WARNING: total FTE estimate %.1f is within %d of the ALE threshold of %d — review ACA Applicable Large Employer status.\n", total_fte, buffer, threshold;
+      }
+    }
+  ' "{{aca_headcount_csv_path}}"
+variables:
+  aca_headcount_csv_path:
+    default: "./aca-headcount-hours.csv"
+    description: "Path to a CSV with columns: employee_name,employment_type (full-time/part-time/variable),monthly_hours (only used for non-full-time rows)."
+  fte_threshold:
+    default: 50
+    description: The ACA Applicable Large Employer FTE threshold to watch for (50 under current federal rules; confirm against current law).
+  warn_buffer:
+    default: 5
+    description: How close to the threshold (in FTE count) triggers a warning.
+compatible_agents: [generic, claude, codex]
+notes: >
+  This is a rough monthly estimate using the standard 120-hours-per-month
+  divisor for part-time/variable-hour employees, not a substitute for the
+  IRS's official monthly measurement method — consult counsel/a benefits
+  advisor before relying on it for actual ALE determination.

package/jobs/hr/cobra-notification-deadline-tracker.yaml

@@ -0,0 +1,47 @@
+id: cobra-notification-deadline-tracker
+version: 1
+name: COBRA Notification Deadline Tracker
+description: >
+  Flags terminated employees who are approaching or past the deadline for
+  their COBRA election notice with no notice sent on record. Use this if
+  a COBRA notice has ever gone out late — or not at all — because
+  offboarding paperwork got deprioritized once the employee's last day
+  had already passed.
+category: hr
+tags: [hr, cobra, offboarding, compliance, benefits]
+schedule: "0 8 * * 1-5"
+timezone: "America/New_York"
+runner: shell
+command: >
+  today=$(date +%s);
+  awk -F',' -v today="$today" -v deadline_days="{{notice_deadline_days}}" -v warn="{{warn_days_before}}" '
+    NR==1 {next}
+    $3 != "" {next}
+    {
+      cmd = "date -d \"" $2 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $2 "\" +%s 2>/dev/null";
+      cmd | getline term_epoch; close(cmd);
+      days_since_term = (today - term_epoch) / 86400;
+      days_left = deadline_days - days_since_term;
+      if (days_left < 0) {
+        printf "COBRA NOTICE OVERDUE: %s — terminated %s, %d day(s) past the %d-day notice deadline, no notice sent\n", $1, $2, -days_left, deadline_days;
+      } else if (days_left <= warn) {
+        printf "COBRA NOTICE DUE: %s — terminated %s, %d day(s) left of the %d-day notice deadline, no notice sent\n", $1, $2, days_left, deadline_days;
+      }
+    }
+  ' "{{cobra_tracking_csv_path}}"
+variables:
+  cobra_tracking_csv_path:
+    default: "./cobra-tracking.csv"
+    description: "Path to a CSV with columns: employee_name,termination_date (YYYY-MM-DD),notice_sent_date (blank if not yet sent)."
+  notice_deadline_days:
+    default: 44
+    description: Days from the qualifying event (termination) to get the COBRA election notice out, per your plan's timeline (federal rules generally allow up to 44 days combined; confirm against your plan administrator's actual deadline).
+  warn_days_before:
+    default: 10
+    description: How many days before the deadline to start flagging as due.
+compatible_agents: [generic, claude, codex]
+notes: >
+  Federal COBRA timelines and state mini-COBRA rules vary — confirm the
+  actual notice_deadline_days with your plan administrator or counsel
+  rather than relying on the default. Assumes an already-exported
+  termination/notice CSV rather than live benefits-system access.

package/jobs/hr/compliance-training-completion-check.yaml

@@ -0,0 +1,41 @@
+id: compliance-training-completion-check
+version: 1
+name: Compliance Training Completion Check
+description: >
+  Flags employees with a required compliance training (harassment
+  prevention, safety cert, data-privacy, etc.) approaching or past its due
+  date with no completion on record. Use this if a training deadline has
+  ever been missed and only surfaced when a regulator or auditor asked for
+  proof of completion.
+category: hr
+tags: [hr, compliance, training, deadline]
+schedule: "0 8 * * 1"
+timezone: "America/New_York"
+runner: shell
+command: >
+  awk -F',' -v days="{{warn_days_before}}" '
+    NR==1 {next}
+    $4 != "" {next}
+    {
+      cmd = "date -d \"" $3 "\" +%s 2>/dev/null || date -j -f %Y-%m-%d \"" $3 "\" +%s 2>/dev/null";
+      cmd | getline due_epoch; close(cmd);
+      "date +%s" | getline now_epoch; close("date +%s");
+      diff_days = (due_epoch - now_epoch) / 86400;
+      if (diff_days <= days) {
+        printf "TRAINING DUE: %s — \"%s\" due %s (%d days), not completed\n", $1, $2, $3, diff_days;
+      }
+    }
+  ' "{{training_tracking_csv_path}}"
+variables:
+  training_tracking_csv_path:
+    default: "./compliance-training-tracking.csv"
+    description: "Path to a CSV with columns: employee_name,training_name,due_date (YYYY-MM-DD),completed_date (blank if not yet done). One row per required training per employee."
+  warn_days_before:
+    default: 14
+    description: How many days before a training's due date to start warning.
+compatible_agents: [generic, claude, codex]
+notes: >
+  Assumes a maintained export from your LMS/HRIS listing one row per
+  employee per required training. Distinct from new-hire-paperwork-check,
+  which covers one-time onboarding documents rather than recurring/annual
+  compliance training.