binky-control/templates/OK-Audit-Template-v1.md
tegwick d2f68f0f53 BINKY-WP-0003-T02: OK audit template v1 (AWQ-001)
Company-agnostic flagship-offer template generalized from the Binky
dogfood run. audit-core assessed as home candidate — it owns audit event
logging, not company audits, so the template stays in binky-control.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-16 10:56:17 +02:00

4.9 KiB
Raw Permalink Blame History

Operational Knowledge Audit — Template v1

Status: v1 — 2026-07-16 (BINKY-WP-0003-T02, AWQ-001). Company-agnostic template generalized from the Binky dogfood run (audits/OK-Audit-v0-binky.md). This is the delivery backbone of the flagship offer in OfferCatalog.md. Home note: the audit-core repo was assessed as a candidate home but owns audit event logging (JSONL backend contract), not company audits — the template lives here until a dedicated offer-delivery repo exists.

What this audit is

A structured inventory of a company's operational knowledge: what exists, where the evidence lives, what's missing, and what to do about it — each action tagged with an owner and an autonomy lane so agents can execute the safe parts immediately.

Deliverables (fixed):

  1. Findings document (this template, filled) — one section per area.
  2. Risk register entries (likelihood / impact / mitigation / escalation date).
  3. Action queues: decisions the founder must make, batched external-contact items ("command day" runbook), and agent-executable backlog.
  4. Machine-readable summary (YAML, below) for tooling.

Method

  • Evidence-first. Every finding cites a source (document, account statement, registry lookup, founder statement). Founder-stated facts are marked as such until verified against records.
  • Gaps are findings. "Unknown / to verify" is a first-class result.
  • Every action gets a lane. Green (agent does it), Blue (agent does it, founder informed), Yellow (agent prepares, founder approves), Orange (founder executes with prepared material), Red (founder only).
  • Time-boxed. v1 targets one working day of agent time plus one founder interview (~60 min). Depth grows in follow-up audits, not this one.

Areas

Registry entry (name, seat, address, officers), articles of association, shareholder structure, company purpose vs. actual/planned activity, pending obligations (transparency register, announcements). Typical gaps: stale registered address, purpose incompatible with new business, missing document archive.

2. Tax & accounting

Accountant mandate status, filing status per year (annual accounts, tax returns, advance VAT), bookkeeping system and access ownership, receipt backlog, special patterns needing professional review (self-billing, intra-owner transactions). Typical gaps: no active mandate, filings behind, bookkeeping access owned by a third party.

3. Banking & finance

Accounts and access, cost base (recurring subscriptions, fees, rent), revenue history and paths, payment infrastructure, cash/runway view. Typical gaps: no consolidated cost table, dormant expensive accounts, no runway visibility.

4. Digital presence & communications

Domains owned, company email (own domain? credential custody? agent integration?), website, paper-mail handling, where official notices actually arrive. Typical gaps: correspondence over private addresses, mail with no intake pipeline, credentials outside any vault.

5. Contracts & IP

Existing contracts and obligations, IP ownership (company vs. founder privately — especially software/repos), evidence trail for commissioned work, template availability (NDA, service agreement). Typical gaps: unclarified IP before productizing, verbal agreements.

6. Risks & queues (synthesis)

Consolidate: new/updated risk entries, decision packages, external-contact batch, agent backlog. Close with a one-paragraph verdict: overall state, the 35 reactivation/operation-critical gaps, and the single best next action.

Per-area record format

For each area record: evidence (with source + confidence), gaps (unknowns count), actions (each: description, owner, lane, queue destination).

Machine-readable summary

audit: <company>
version: 1
date: ""
auditor: ""            # agent/person
founder_interview: ""  # date or n/a
areas:
  legal_registry:    {evidence: [], gaps: [], actions: []}
  tax_accounting:    {evidence: [], gaps: [], actions: []}
  banking_finance:   {evidence: [], gaps: [], actions: []}
  digital_presence:  {evidence: [], gaps: [], actions: []}
  contracts_ip:      {evidence: [], gaps: [], actions: []}
risks: []              # ids of created/updated risk entries
queues:
  decisions: []        # decision-package ids
  external_batch: []   # command-day items
  agent_backlog: []    # autopilot items
verdict: ""
# action item shape:
#   {desc: "", owner: founder|accountant|bank|agent, lane: green|blue|yellow|orange|red, queue: ""}
# evidence item shape:
#   {fact: "", source: "", confidence: stated|documented|verified}

Evidence from the dogfood run

Binky Hedgehog GmbH ran this audit on itself first (audits/OK-Audit-v0-binky.md): it surfaced a missing accountant mandate, a bookkeeping-access dependency, and an unclarified IP position — and turned them into queued, lane-tagged actions within the same day. That run is the reference example for offer delivery.