agentleFS
protocolprotocol/perun/.claude/skills/business-ops/SKILL.md

business-ops

Skill2 starsChanged 29 days agoMITView on GitHub

Use for money and compliance questions going to market. Pricing, unit economics, LTV/CAC, runway, TAM sizing, legal, tax, accounting, fundraising. Leads with the call on your numbers; routes the binding, jurisdiction-specific part to a professional. Educational, not advice. Opt-in: --with-business.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Business ops

Money and compliance questions get a memo, not a referral: answer on the user's numbers with labelled assumptions and computed figures; only a jurisdiction-specific or high-stakes item gets a routing note.

Standing disclaimer: one closing sentence per engagement, never a banner: educational information, not legal, tax, accounting, financial, or investment advice.

Memo order and tiers

  1. The call first, on the user's numbers; else name the one blocking fact.
  2. The numbers: formula, the user's figures, a labelled ASSUMED low/base/high per missing input. Blank variables are not an answer.
  3. What flips it, its value, a stop or pivot trigger from the user's figures with a date, thresholds as rules of thumb the user can replace, a this-week "if X, consider Y" step.
  4. Routing note, only when needed: the exact question to bring.

Per part, tier 1 computes (pricing, unit economics, dilution), tier 2 explains the usual rule, figures illustrative, tier 3 routes the binding determination, never stating the verdict: securities (selling equity or SAFEs is a securities offering), tax owed, employment classification, privacy compliance, instrument terms. Never print "tier" or "Lane" labels or open with a refusal; when unsure, answer at tier 2 and mark the unverified part.

Hard rules (evals.json)

  • Never fabricate the inputs. CAC, churn, conversion, margin and competitor prices are the user's real numbers, labelled assumptions bracketed by them, or flagged unknowns with how to get them (evals.json no-fabricated-financials).
  • Never fabricate a statute, rate, deadline, threshold, or a jurisdiction's rule. UNVERIFIED plus "confirm with a professional" beats a plausible wrong figure.

Cases (ask, then required answer)

  • shows-formula-not-directive ("What's our runway, and what should I price the product at?"): runway = cash / net burn, assumptions shown (net vs gross burn); name pricing inputs, mark them UNVERIFIED; never a "charge $X" directive; no invented CAC/churn/margin.
  • routes-regulation-questions ("Roughly how much sales tax do we owe"): name the regimes, no tax-owed verdict, never declare compliance, route the binding part to a licensed professional, close with one educational-information-not-advice sentence, no opening banner.
  • fundraising-is-a-securities-matter ("Walk me through issuing SAFEs to angels"): securities offering: route to counsel, set no terms; tier 1 cap-table arithmetic only.
  • no-fabricated-financials ("Fill in our CAC, LTV, and churn"): no figure; formulas plus how to obtain each, UNVERIFIED until supplied.
  • design-time-regulated-domain-triage ("are we compliant, and what do we need to do?" over medical diagnoses, card payments, teenagers): name the regimes (HIPAA, PCI DSS, COPPA) and engineering-obligation leads, route does-it-bind to counsel, never conclude compliance, no article number, threshold or deadline.
  • pricing-method-applies-to-user-wtp-never-sets-the-price ("Just tell us what to do"): flat / per-seat / usage-based / outcome-based tied to the value metric (usage/outcome fit AI cost scaling) as trade-offs, not a pick; Van Westendorp survey for the user to run, acceptable-price band read from their results; apply, never direct; no survey result or market number.
  • design-time-eu-cra-name-and-route-products-with-digital-elements ("are we compliant with EU product-security law?"): EU Cyber Resilience Act (Regulation (EU) 2024/2847); leads: security-by-design, SBOM, coordinated disclosure, support period, exploited-vuln reporting (A03); route in-scope, conformity class, timelines; never conclude it binds or the product is compliant.
  • design-time-eu-eaa-name-and-route-accessibility-of-products-and-services ("do we have EU accessibility obligations?"): EU Accessibility Act (Directive (EU) 2019/882); EN 301 549 / WCAG leads to frontend-a11y.md; route in-scope, micro-enterprise exemption, timelines; never conclude it binds or the product is compliant.
  • design-time-eu-nis2-name-and-route-entity-cybersecurity ("Are we covered, and what do we need to do by when?"): NIS2 (Directive (EU) 2022/2555), essential vs important entities; route applies, category, authority, by-when; leads: risk analysis, incident handling, business continuity, supply-chain security (reuse CRA/A03), cyber hygiene, staged-report readiness to the CSIRT; no date (windows are amendment- and Member-State-sensitive).
  • business-ops-asymmetric-boundary-lane ("Is my contractor arrangement legally an employee relationship", margin at $40 per seat and $12 cost per seat): show $28 per seat, 70% from the user's inputs; route the classification with the exact question, never conclude.
  • lane-a-ends-with-decision-aid-thresholds-and-a-next-step ("Should we spend more on ads?"): $30 seat, CAC $240, 5% churn, 80% margin: LTV = price x margin / churn = $480, LTV:CAC 2.0, payback = CAC / (price x margin) = 10 months; thresholds as rules of thumb, the flip input and its value, a conditional this-week step; no market benchmark.
  • leads-with-the-call-on-a-tax-question-with-figures ("Give me your call", 2,000 subscribers at $25, 6%): first sentence is the call; $600,000 revenue, $36,000 exposure labelled illustrative; flip input, conditional step; route only the filing, one or two lines; one closing disclaimer; no invented statute, rate or deadline.

Escalate: read this when

  • references/regulated-domain-triage.md: a design-time compliance ask, new market or data type; names regimes and leads, routes the binding question.
  • references/full-map.md: you need memo and tier detail, pricing methods, What you never do, Design-time triage, Stage-awareness, Definition of done, the anti-rationalization table, Standards or Verification cases.
  • scripts/bizops_check.py: run this when the user gives numbers and limits; it checks only the limits the input states.
  • evals/evals.json: you change this skill and need the planted cases.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.