Index of every page on this site: llms.txt
All skills
jeremylongshore avatar

/snowflake-strong-auth-migration-pilot

@820dfc8
Run it yourself

Pilot a Snowflake authentication modernization without lockouts or credential exposure. Inventory PERSON, SERVICE, and LEGACY_SERVICE users; map named workloads to WIF, key pair, OAuth, or bounded PAT; and audit managed MCP/OAuth session controls. Use when replacing Snowflake passwords, reviewing service identities, planning workload identity federation, or scoping MCP OAuth. Trigger with phrases like "Snowflake auth migration", "service user password", "Snowflake WIF", "key pair rotation", or "MCP OAuth role scope".

Use this Skill: https://skilld.dev/gh/jeremylongshore/tons-of-skills-marketplace/snowflake-strong-auth-migration-pilot

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

referencescurrent-evidence-contract.md

≈2.6k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Current and Historical Evidence Contract

Use this contract when collecting or validating strong-auth migration evidence. It defines a bounded, read-only evidence lane. It does not grant permission to change users, credentials, integrations, authentication policies, or sessions.

Evidence classes

The analyzer requires three independent live collector receipts:

Collection key Surface Source Dataset What it supports
current auth-current SHOW USERS current_users Near-current configuration posture, subject to active-role visibility
historical auth SNOWFLAKE.ACCOUNT_USAGE.USERS historical_users Delayed configuration corroboration
login_history auth-login-history SNOWFLAKE.ACCOUNT_USAGE.LOGIN_HISTORY login_history Settled portion of the trailing seven-day authentication horizon

Each receipt also contains exactly one same-statement execution_context row. The organization-plus-account identifier, collector user, primary role, and secondary-role representation are SHA-256 pseudonyms. The analyzer requires equivalent authorization context across all three invocations but does not claim they used one physical session. The account pseudonym hashes CURRENT_ORGANIZATION_NAME() plus CURRENT_ACCOUNT_NAME(); the reusable legacy locator is not the binding key. primary_role_type accepts only Snowflake's documented ROLE and APPLICATION_INSTANCE values.

The reviewed LOGIN_HISTORY SQL deliberately excludes the newest two hours, so its effective settled interval is the older portion of the trailing seven-day horizon rather than seven full settled days. Both Account Usage sources can lag by up to 120 minutes. A present event is an observation. An absent event does not prove that authentication never occurred.

Privacy boundary

Permitted identity correlation is user_name_sha256, computed inside Snowflake from the exact username representation returned by each source. SHA-256 is pseudonymization, not anonymity; common names remain guessable.

Receipt rows must not contain raw usernames, login/display/profile names, email, client IP, connection or private-link identifiers, raw event or factor IDs, client version, login details, free-form error messages, credential comments, PAT names, WIF issuer/subject/audience data, public-key fingerprints, secrets, tokens, passwords, or private keys. Unknown fields fail closed rather than being silently copied into the report.

REPORTED_CLIENT_TYPE is excluded entirely because Snowflake documents it as unauthenticated telemetry. Snowflake-managed SNOWFLAKE_SERVICE principals are retained in both raw user datasets so a 10,000-row cap cannot be hidden, marked SNOWFLAKE_MANAGED_EXCLUDED, and then excluded from the operator denominator. Operator-owned SERVICE_AGENT principals remain in scope and are treated as service identities. A source TYPE=NULL is normalized to Snowflake's documented PERSON meaning.

For operator-owned SERVICE and SERVICE_AGENT principals, Snowflake may expose the same non-applicable password or MFA field as FALSE in SHOW USERS and NULL in Account Usage. Reconciliation normalizes both to the same negative observation; neither form is treated as positive posture evidence.

Bundle envelope

Combine the three unchanged collector receipts with an operator-owned workload inventory. metadata.coverage.user_name_sha256 is the explicit denominator and must match the current receipt and the operator user mapping exactly.

{
  "schema_version": "2.0",
  "metadata": {
    "evaluated_at": "2026-09-03T12:05:00Z",
    "max_age_seconds": 3600,
    "connection_profile": "auth-readonly",
    "login_history_latency_seconds": 7200,
    "authorization_context": {
      "account_identifier_sha256": "<64-lowercase-hex>",
      "collector_user_sha256": "<64-lowercase-hex>",
      "primary_role_sha256": "<64-lowercase-hex>",
      "primary_role_type": "ROLE",
      "secondary_roles_sha256": "<64-lowercase-hex>"
    },
    "coverage": {
      "user_name_sha256": ["<64-lowercase-hex>"]
    }
  },
  "collections": {
    "current": {"receipt": {}},
    "historical": {"receipt": {}},
    "login_history": {"receipt": {}}
  },
  "users": [
    {
      "name": "ETL_SVC",
      "user_name_sha256": "<same-64-lowercase-hex>",
      "type": "LEGACY_SERVICE",
      "auth_methods": ["PASSWORD", "WIF"],
      "owner": "data-platform"
    }
  ],
  "workloads": [
    {
      "name": "ETL_PROD",
      "identity": "ETL_SVC",
      "identity_sha256": "<same-64-lowercase-hex>",
      "owner": "data-platform",
      "current_auth": "PASSWORD",
      "supported_auth": ["WIF", "KEY_PAIR"],
      "roles": ["ETL_ROLE"]
    }
  ],
  "integrations": [],
  "enforcement_windows": [
    {
      "name": "etl-pilot",
      "workload": "ETL_PROD",
      "identity_sha256": "<same-64-lowercase-hex>",
      "target_auth": "WIF",
      "start": "2026-09-03T08:00:00Z",
      "end": "2026-09-03T09:00:00Z",
      "owner": "data-platform",
      "approved_by": "security-approver",
      "change_id": "CHG-1001"
    }
  ]
}

Use stable internal workload and change references; do not put ticket prose or secrets in receipt rows. Raw names in the separate owner inventory are local operator inputs and are never inferred from pseudonyms or emitted as collector evidence. Operator type, auth_methods, and workload current_auth values must reconcile to current receipted posture. Methods that user posture cannot prove, such as OAuth or SAML, do not support scoped completeness in this bundle. MFA is a separate factor posture flag, not a primary auth_methods value. Workload names are unique; duplicates cannot collapse the enforcement-window denominator. The top-level bundle, metadata object, receipt wrapper, receipt, datasets, and projected rows are exact schemas: unknown fields fail closed. A connection profile is only a local profile name containing letters, digits, dot, underscore, or hyphen; it is never a connection string or credential field.

Canary and break-glass evidence is deliberately not part of this schema. Keep those operational receipts in the separately controlled approval workflow; embedding either object is rejected so arbitrary telemetry cannot be echoed or mistaken for independently verified proof.

Trust and freshness

The collector's receipt_sha256 detects accidental receipt changes but can be recomputed by anyone who can edit the file. Before analysis, compute the canonical whole-bundle digest with --print-input-sha256 and retain it outside the bundle at a controlled local boundary. Supply that value through --trusted-input-sha256.

Only DIGEST_MATCHED_OPERATOR_ASSERTED permits receipt datasets into scoped reconciliation. This status means byte identity with the separately recorded bundle; it is not a signature, origin attestation, or statement about who collected it.

metadata.evaluated_at must be within five minutes of the analyzer's actual UTC clock, and metadata.max_age_seconds cannot exceed 3600. This prevents a caller from moving the entire evidence timeline backward or declaring arbitrarily old receipts fresh. Every receipt must be live, recent, internally ordered, below its reviewed cap, and bound to the exact bundled SQL, canonical nonclaims, expected authorization context, and expected source/dataset fields. Offline, stale, future-dated, errored, capped, privilege-filtered, or context-mismatched receipts are quarantined. Freshness is measured from each same-statement execution_context.observed_at, not from later CLI completion, and the entire collection interval must fit inside the declared maximum age.

Reconciliation and claims

The current and historical rows join only on user_name_sha256. The analyzer also compares normalized created_on to detect same-name principal recreation, then compares disabled, type, and the password, RSA, MFA, PAT, and workload-identity posture flags. SERVICE and SERVICE_AGENT password/MFA fields may be NULL only where Snowflake defines them as non-applicable; other unknown posture blocks completeness. A login observation older than the reconciled created_on is not attributed to the current principal. Current-only, historical-only, duplicate, malformed, or field-drift rows require review; delayed history never overrides the current SHOW observation.

evidence_scope_complete is limited to the declared pseudonymous denominator. It is not account-wide proof. LOGIN_HISTORY cannot set canary, recovery, or cutover readiness by itself. Positive target login/action, negative old-path and scope outcomes, and a separately tested recovery path remain human approval gates. Likewise, supported_auth is an operator declaration, not independent proof that the runtime, driver, connector, or Snowflake integration supports the selected target. Current posture flags describe current configuration only.

The analyzer performs no Snowflake operation, and the reviewed collector SQL is read-only. Neither component attests to actions taken elsewhere in the surrounding session or workflow.

Least-privilege and source limitations

Use an established read-only profile. SNOWFLAKE.SECURITY_VIEWER is the documented database role for the Account Usage security views, but it is broader than these two queries. SHOW USERS exposes detailed columns only when the active role has the documented visibility; this workflow reports the gap and does not grant MANAGE GRANTS or any other privilege.

Snowflake references:

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No alerts29d3 checks · Risk SAFE
  • Gen Agent Trust Hub29d

    This skill is a security tool designed to pilot Snowflake authentication migrations. It focuses on read-only evidence collection and planning, featuring extensive built-in sanitization to prevent credential exposure. No malicious patterns, prompt injections, or data exfiltration vectors were detected.

  • Socket1mo

    No alerts

  • Snyk1mo

    Risk: LOW · No issues

Signed by skilld at 820dfc8. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated 3 months ago
What it can do
Reads files Runs commands
argument-hint
[redacted-auth-evidence.json]
version
3.16.0
author
Jeremy Longshore <jeremy@intentsolutions.io>
All 2 allowed tools
ReadBash(python3:*)
Other metadata
compatibility
Model-agnostic workflow; requires Python 3.10+; optional Snowflake CLI for live read-only evidence collection
tags
[
  "saas",
  "snowflake",
  "security",
  "authentication",
  "wif",
  "oauth",
  "mcp"
]

README badge

README badge for jeremylongshore/tons-of-skills-marketplace/snowflake-strong-auth-migration-pilot