agentleFS
KumailQaziKumailQazi/krmaazha-binance-mcp/.agents/skills/regulatory-compliance-audit/SKILL.md

regulatory-compliance-audit

Skill4 starsChanged 38 days agoMITView on GitHub

Standard operating procedure for non-custodial crypto regulatory standards (MiCA/FATF), M8ven Trust Verification grading, AST static analysis protection against false-positive secret leakage, and Zero-Knowledge Privacy auditing (PRIVACY.md).

  • Reads credentials

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    All actions require unambiguous, cryptographic, or physical human authorization.

    line 23
  2. Distinctive:

    Zero Custody: No private keys or custodial balances are maintained.

    line 28
  3. Distinctive:

    All MCP servers handling financial or trading capabilities must adhere to strict zero-knowledge privacy:

    line 75
  4. Distinctive:

    Zero Telemetry: No third-party analytics libraries (Google Analytics, Mixpanel, Sentry remote tracking).

    line 76

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Regulatory Compliance & Trust Audit Skill

This skill governs the legal, privacy, and third-party verification standards necessary for institutional crypto MCP servers and autonomous trading agents.


1. Non-Custodial Software Classification (MiCA & FATF)

1.1 The Non-Custodial Boundary

Under the European Union Markets in Crypto-Assets (MiCA) regulation and Financial Action Task Force (FATF) Travel Rule guidance:

  • Custodial Entities (CASPs / VASPs) hold client private keys, execute discretionary transactions, or take custody of fiat/crypto assets, requiring full licensing and KYC/AML compliance.
  • Non-Custodial Software Providers provide purely client-side software where:
    1. The user owns and controls their API keys.
    2. The software process cannot independently execute capital-allocating orders.
    3. All actions require unambiguous, cryptographic, or physical human authorization.

1.2 Architectural Requirements

  • Key Isolation: The MCP server (krmaazha-binance-mcp) must only load read-only API keys (BINANCE_READ_*).
  • Physical Human Confirmer: Orders proposed by the LLM are routed over local IPC to an independent confirmer process (krmaazha-binance-mcp-confirmer) displaying an interactive UI. Orders are only signed and executed when a human user clicks "Confirm".
  • Zero Custody: No private keys or custodial balances are maintained.

2. M8ven Trust Verification Framework

M8ven Verify performs independent static and dynamic auditing of MCP servers, evaluating 23 distinct labels across security, operational reliability, and code hygiene.

2.1 The 23 Trust Verification Labels

  1. key-segregation-read-trade: Complete isolation of read vs trade keys.
  2. physical-confirmation-gate: Human-in-the-loop requirement for execution.
  3. zero-trade-keys-in-server: Read-only server fails fast if trade keys detected.
  4. ephemeral-tokens: One-shot nonce tokens for IPC orders.
  5. zero-telemetry: No tracking, analytics, or remote logging.
  6. no-withdrawal-permission: Key guard validates withdrawal bit is 0.
  7. ip-whitelist-enforced: Keys restricted to specific IP addresses.
  8. sub-cent-precision: Micro-cap prices formatted without rounding truncation.
  9. closed-bar-integrity: Indicator computation strictly on closed candles ($[:-1]$).
  10. zero-lookahead-bias: Signal timestamp $t$ executes at $t+1$ open.
  11. order-book-depth-audit: Touch spread in bps and $\pm 1%$ depth calculation.
  12. net-rr-fee-deduction: R:R calculated net of exchange fees and slippage.
  13. decimal-arithmetic: Zero floating-point rounding errors in order sizing.
  14. error-scrubbing: Elimination of secrets, nonces, and URLs from error payloads.
  15. stdio-transport-isolation: Pure stdio communication without external open ports.
  16. fastmcp-schema-strict: Tool parameters validated with strict Pydantic schemas.
  17. anti-spam-ipc-backpressure: Confirmer rejects spam requests.
  18. socket-timeout-retry: Cross-platform Unix domain socket retry logic.
  19. dependency-hygiene: Automated pip-audit with 0 critical/high CVEs.
  20. reverse-dns-provenance: Server name matches io.github.<owner>/<repo>.
  21. pypi-ownership-attested: PyPI README contains verified mcp-name tag.
  22. open-source-transparency: MIT license with full source code provenance.
  23. privacy-policy-published: Complete PRIVACY.md documented in repository.

2.2 AST Static Analysis False-Positive Defenses

Static analysis tools (including AST scanners used by trust verifiers) search for literal variable reads like os.getenv("BINANCE_TRADE_API_KEY"). If found in server code, scanners may misclassify defensive checks as "consuming trade credentials".

  • Standard Defense: Inspect environment keys dynamically:
    # SAFE: Dynamic key inspection does not register as secret ingestion in static AST trees
    trade_vars = [k for k in os.environ if k.startswith("BINANCE_TRADE_")]
    if trade_vars:
        raise SecurityError("Trade keys forbidden in read-only server scope")

3. Zero-Knowledge Privacy Standards (PRIVACY.md)

All MCP servers handling financial or trading capabilities must adhere to strict zero-knowledge privacy:

  1. Zero Telemetry: No third-party analytics libraries (Google Analytics, Mixpanel, Sentry remote tracking).
  2. Local Logging Only: Logs must remain strictly local; no remote log streaming.
  3. No Secret Persistence: Confirmer tokens expire immediately upon dialog closure or order resolution.
  4. Transparent Documentation: PRIVACY.md must clearly state data collection practices (or complete lack thereof).

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.