ar-collections-agent

Automate AR collections with aging analysis, prioritization, email drafts, DSO reporting, payment tracking, and bad debt reserve recommendations.

Security Scan

The skill's requirements and instructions are coherent with an AR collections assistant: it is instruction-only, asks for no unrelated credentials or installs, and contains reasonable guardrails — no clear mismatches were found.

Purpose & Capability

Name, description, and SKILL.md all describe AR aging, prioritization, email drafts, DSO/KPI calculations and bad-debt recommendations; required items (none) are proportionate to an instruction-only analyzer that works from CSV/exports or another QBO integration.

Instruction Scope

Instructions stay focused on AR analysis, drafting, and reporting and include explicit guardrails (read-only by default, approval required for sending emails, no legal actions). Note: it references 'direct QBO API' and integration with a separate 'qbo-automation' skill but does not describe authentication flows — that dependency should supply credentials. It also mentions persisting a tracker at memory/ar-promise-tracker.md if requested; verify where this memory is stored and its access controls.

Install Mechanism

Instruction-only skill with no install spec and no code files — minimal install risk (nothing is downloaded or written by the skill itself).

Credentials

The skill declares no required environment variables or credentials, which is fine for CSV-based use. However, the declared capability to operate via the QBO API implies that QBO credentials would be needed when that integration is used; those creds are not declared here and should be provided only via a trusted integration or separate skill (e.g., qbo-automation).

Persistence & Privilege

Does not request 'always' or other expanded platform privileges. It may persist a simple tracker file to agent memory when explicitly requested; this is limited and documented in the SKILL.md, but confirm storage location and retention policy before enabling persistence.

Assessment

This is an instruction-only AR collections helper that appears coherent and low-risk, but check these before installing: (1) Confirm how QBO integration is handled — QBO credentials should come from a trusted integration and only be granted if needed. (2) Require explicit human approval before sending emails or performing writes; the SKILL.md already says that, but enforce it in your agent settings. (3) Verify where memory/ar-promise-tracker.md will be stored and who can read it (sensitive invoice/customer data). (4) Review email templates and legal guardrails so Tier 4/5 messages aren’t sent unintentionally. (5) If you plan to use direct QBO API features, audit the qbo-automation skill or connector that will supply credentials. No code is installed by this skill, and the scanner found no code-based signals.

Automate and systematize accounts receivable collections: aging analysis, prioritization, outreach drafts, DSO tracking, and bad debt analysis.

Inputs Accepted

• QBO AR Aging Summary/Detail (CSV or PDF export)
• Invoice CSV: invoice_id, customer, invoice_date, due_date, amount, amount_due, status
• Direct QBO query results (via qbo-automation skill)
• Manually pasted invoice data

Core Workflows

1. AR Aging Analysis

Bucket invoices into standard aging tiers:

Bucket	Days Past Due
Current	0
1–30	1–30
31–60	31–60
61–90	61–90
91–120	91–120
120+	121+

Calculate per customer:

• Total AR balance
• Oldest invoice date
• % of balance in each bucket
• Days Sales Outstanding (DSO) = (AR Balance / Total Credit Sales) × Days in Period

Flag high-risk accounts: balance > $5k AND 60+ days past due, or any invoice 120+ days.

2. Prioritization Matrix

Score each overdue account (1–10):

• Balance weight (40%): Higher balance = higher score
• Age weight (40%): Older = higher score
• History weight (20%): Prior payment promises broken = higher score

Output: ranked list, top 10 accounts to contact first.

3. Collection Outreach Drafts

Draft escalating messages based on aging tier. See references/email-templates.md for full templates.

Tone ladder:

• 1–30 days: Friendly reminder
• 31–60 days: Second notice, reference invoice
• 61–90 days: Firm request, payment plan offer
• 91–120 days: Final notice, escalation warning
• 120+ days: Collections/legal escalation notice

Always include: invoice numbers, amounts, due dates, payment link/instructions.

Never send without Irfan's approval. Draft only unless explicitly authorized.

4. DSO & KPI Dashboard

Calculate and report:

• DSO (overall and per customer segment)
• Collection Effectiveness Index (CEI): (Beginning AR + Credit Sales − Ending AR) / (Beginning AR + Credit Sales − Current AR) × 100
• Bad Debt %: Write-offs / Total Credit Sales × 100
• % AR Current vs. Aged

Output format: summary table + bullet insights for client presentation.

5. Bad Debt Reserve & Write-Off Analysis

Apply allowance method by aging bucket (customize per client):

Bucket	Default Reserve %
Current	0.5%
1–30	2%
31–60	5%
61–90	15%
91–120	30%
120+	50–100%

Output: recommended reserve amount, GL journal entry draft, write-off candidates.

6. Payment Promise Tracker

When a customer commits to pay by a date:

1. Log: customer, amount, promise date, follow-up date (3 days before)
2. Flag broken promises for escalation tier bump
3. Output: follow-up schedule CSV or table

Save tracker state to memory/ar-promise-tracker.md if persistence requested.

Output Formats

• AR Aging Report: Markdown table or CSV
• Collection Priority List: Ranked table with scores
• Email Drafts: Ready-to-send (with approval), parameterized per customer
• DSO Dashboard: Summary metrics + trend commentary
• Bad Debt Schedule: GL-ready journal entries
• Follow-Up Schedule: Date-indexed action list

Reference Files

• references/email-templates.md — Full escalation email templates (all 5 tiers)
• references/ar-kpi-formulas.md — DSO, CEI, turnover ratio formulas with examples

Guardrails

• Read-only default: Never modify QBO or send emails without explicit approval
• No legal advice: Flag 120+ day accounts for attorney review; do not draft demand letters
• Client data isolation: Never mix AR data across clients in the same analysis
• Write-off authority: Recommend only; write-off execution requires Irfan sign-off
• PTIN boundary: Bad debt tax deduction guidance requires PTIN-backed service; flag and refer