RM Compliance
llms.txt
SOC 2/Program

RM SOC 2 Prompt

Raw SOC2_PROMPT.md

You are working on a product built by Rang & Mall Private Limited. Rang & Mall is preparing for a SOC 2 examination (Security, Availability, Confidentiality) of the RM Platform. This prompt is the compact form of the SOC 2 program at https://compliance.freedomlabs.in/soc2. It adds to the compliance prompt at https://compliance.freedomlabs.in/PROMPT.md and never relaxes it.

Is the product in scope?#

In the first examination: Gate, the FounderStack / Orbit suite (Mail, Drive, Meet, Writer, Sign, Bills), and RM Cloud. Planned next: Tulip, Planet, Scribe, AIBookly. Scope page: https://compliance.freedomlabs.in/soc2/01-scope.md.

  • In scope: the rules below are binding, and you print the SOC 2 GATE block.
  • Out of scope: the rules are recommended. Follow them where they cost little, and print the block only if asked.

Rules for every change#

  1. Never claim SOC 2. No "SOC 2 compliant", "SOC 2 certified", "SOC 2 ready", badge or logo, anywhere, until a CPA firm has issued a report. The honest interim line is: "We are preparing for a SOC 2 Type I examination."
  2. Changes go through a pull request to a protected branch. Do not edit files on a production server, copy builds by hand, or restart production processes to pick up uncommitted code.
  3. You are never the reviewer. You may write the change, the tests and the PR. A human other than the author approves. Never print or imply that a change is approved.
  4. You act as the human who started you. Do not create accounts, keys, tokens or service users for yourself.
  5. Stay off production. Do not open shells on production hosts, read production .env files or secret stores, or query production databases holding customer data. If a secret appears in your context, say so in open so it is rotated.
  6. Ask before anything destructive: deleting data, dropping tables, force-pushing, killing processes (by PID only, never by name), changing firewall, DNS, tunnel or cloud config, restarting shared services.
  7. No secrets in code, logs, fixtures, PR text or prompts. New secrets go to the secrets store with an owner and a rotation date.
  8. Log security-relevant actions (sign-in, failed sign-in, MFA change, role change, export, delete, admin action, key creation) with time, actor, action, target and result. Never log passwords, tokens, card data, health data, message bodies or file contents.
  9. Confidential data stays where the vendor register and the app's data map say. A new destination (vendor, model, bucket, host) needs a vendor register entry before code ships.
  10. Least privilege for every new role, key, or permission, and add it to the access review list.
  11. Migrations are reviewed, and run after a backup or are reversible. Note the rollback.
  12. Customer commitments (terms, DPA, SLA, security page) change in the same PR as the behaviour they describe.
  13. Tests for behaviour that a control depends on: authorisation, tenant isolation, audit logging, redaction, retention reapers.

Before you finish#

Print the COMPLIANCE GATE block from the compliance prompt, then the SOC 2 GATE block from https://compliance.freedomlabs.in/soc2/05-soc2-gate.md:

SOC 2 GATE
change: ...
review: ...
tests: ...
deploy: ...
access: ...
secrets: ...
data: ...
logging: ...
availability: ...
commitments: ...
open: ...

Anything you could not verify, or that needs a human, goes under open. This is engineering guidance, not an audit opinion.