@mailrith/agent-skill
Advanced tools
@mailrith/agent-skill - npm Package Compare versions
+4
-2
| # Changelog | ||
| ## 0.1.0-beta.1 | ||
| ## 0.1.0 - 2026-07-24 | ||
| - Keep the read-only Codex and Claude starters within the default OAuth permissions and explain that only workspace changes produce Activity records. | ||
| - Align workflows with resource-bound diagnostics, Subscriber subscription permissions, and Webhooks Read or Write. | ||
| - Add the first installable Mailrith email marketing skill. | ||
| - Add safe subscriber, draft, preview, approval, execution, progress, and diagnostic workflows. | ||
| - Add safe Subscriber, draft, preflight, scoped execution, progress, and diagnostic workflows. | ||
| - Add current connector templates for OpenAI, Claude, Codex, n8n, and Pipedream. |
@@ -30,6 +30,3 @@ { | ||
| "workspace_get": { "enabled": true }, | ||
| "agent_activity_list": { "enabled": true }, | ||
| "agent_activity_get": { "enabled": true }, | ||
| "subscribers_list": { "enabled": true }, | ||
| "broadcasts_get_send_progress": { "enabled": true } | ||
| "subscribers_list": { "enabled": true } | ||
| } | ||
@@ -36,0 +33,0 @@ } |
@@ -6,3 +6,3 @@ [mcp_servers.mailrith] | ||
| required = false | ||
| default_tools_approval_mode = "writes" | ||
| default_tools_approval_mode = "never" | ||
| startup_timeout_sec = 20 | ||
@@ -14,6 +14,3 @@ tool_timeout_sec = 60 | ||
| "workspace_get", | ||
| "agent_activity_list", | ||
| "agent_activity_get", | ||
| "subscribers_list", | ||
| "broadcasts_get_send_progress", | ||
| ] | ||
@@ -20,0 +17,0 @@ |
@@ -8,3 +8,3 @@ { | ||
| "server_label": "mailrith", | ||
| "server_description": "Mailrith email marketing tools with scoped access, action previews, approvals, and activity records.", | ||
| "server_description": "Mailrith email marketing tools with scoped access, sandbox testing, and redacted activity records.", | ||
| "server_url": "https://api.mailrith.com/mcp", | ||
@@ -15,5 +15,5 @@ "authorization": "${MAILRITH_ACCESS_TOKEN}", | ||
| }, | ||
| "require_approval": "always" | ||
| "require_approval": "never" | ||
| } | ||
| ] | ||
| } |
@@ -7,3 +7,3 @@ # Mailrith Connector Templates | ||
| Resolve `${MAILRITH_ACCESS_TOKEN}` in server-side code before sending the request. The starter template exposes read-only tools and keeps provider-level approval on every MCP call. Expand the allowlist only for the workflow you have reviewed. | ||
| Resolve `${MAILRITH_ACCESS_TOKEN}` in server-side code before sending the request. The starter template exposes only read-only tools. Expand the allowlist only for the workflow and Mailrith permissions you intend to grant. | ||
@@ -10,0 +10,0 @@ ## Claude |
@@ -8,3 +8,3 @@ # Mailrith Safety And Retry Rules | ||
| - Inspect resource or activity state before retrying resource-state operations. | ||
| - Do not retry non-idempotent or approval-token operations automatically. | ||
| - Do not retry non-idempotent operations automatically. | ||
| - Treat a timeout after request transmission as uncertain. Look up progress or activity by request ID before doing anything else. | ||
@@ -14,3 +14,3 @@ | ||
| Keep API keys, access tokens, refresh tokens, and approval tokens in the client secret store or process environment. Redact them from errors and tool output. Never place them in email content, Subscriber fields, logs, screenshots, or activity annotations. | ||
| Keep API keys, access tokens, and refresh tokens in the client secret store or process environment. Redact them from errors and tool output. Never place them in email content, Subscriber fields, logs, screenshots, or activity annotations. | ||
@@ -23,2 +23,2 @@ ## Scope Selection | ||
| Before a send, confirm the selected workspace, sender identity, Subscriber filters, exclusions, estimated recipient count, subject, content, and unsubscribe behavior. If any value differs from the approved plan, stop and preview again. | ||
| Before a send, confirm the selected workspace, sender identity, Subscriber filters, exclusions, estimated recipient count, subject, content, and unsubscribe behavior. Run Broadcast preflight again after any relevant value changes. |
@@ -15,7 +15,7 @@ # Mailrith Workflow Recipes | ||
| 3. Use the Subscriber upsert operation for one record or the bulk-import job for a bounded file. | ||
| 4. Preview high-impact eligibility, targeting, or Sequence-enrollment changes. | ||
| 4. Confirm the intended eligibility, targeting, or Sequence-enrollment change before calling the scoped mutation. | ||
| 5. Use a stable source identifier and an idempotency key for retriable writes. | ||
| 6. Return created, updated, skipped, and failed counts without echoing full profiles. | ||
| ## Campaign Draft | ||
| ## Broadcast Preparation | ||
@@ -25,13 +25,20 @@ 1. Read the available Sender Identities, Subscriber targeting options, and templates. | ||
| 3. Confirm subject, sender, recipients, exclusions, and unsubscribe behavior. | ||
| 4. Preview the send operation separately. Drafting permission does not imply sending permission. | ||
| 4. Run Broadcast preflight separately. Drafting permission does not imply sending permission. | ||
| ## Preview And Approved Execution | ||
| ## Sequence Preparation And Operations | ||
| 1. Call the intended operation in plan mode using the final canonical input. | ||
| 2. Show the action ID, risk, target resources, estimated effects, warnings, and expiration. | ||
| 3. Ask the user to review the action in Mailrith under **Settings → Agents → Approvals**. | ||
| 4. Wait until Mailrith reports `approved`. | ||
| 5. Claim the short-lived approval token only in memory and execute the exact planned operation once. | ||
| 6. If Mailrith reports a conflict or changed resource version, create a new plan. Never force the stale plan. | ||
| 1. Use Sequence Preparation to create, read, change, or delete paused Sequences. | ||
| 2. Read sender settings, steps, delays, and current Subscriber count before activation. | ||
| 3. Use a separate Sequence Operations credential to activate, pause, or change individual Subscriber enrollment. | ||
| 4. Treat activation and enrollment as live email operations. | ||
| 5. Confirm the saved status and Subscriber count after each operation. | ||
| ## Preflight And Scoped Execution | ||
| 1. Confirm the credential has the exact execute, bulk, delete, or admin scope required by the operation. | ||
| 2. For Broadcast sends, run preflight and resolve every blocking issue. | ||
| 3. Call the final operation with canonical input and an idempotency key when supported. | ||
| 4. If Mailrith reports a conflict, read the current resource state before deciding whether to retry. | ||
| 5. Use Agent Activity to correlate the request and confirm the result. | ||
| ## Progress And Activity | ||
@@ -41,3 +48,3 @@ | ||
| 2. Use bounded polling and increase the interval for long-running work. | ||
| 3. Read agent activity by request ID, action ID, operation, or resource ID when diagnosing. | ||
| 3. Read Agent Activity by request ID, activity ID, operation, or resource ID when diagnosing. | ||
| 4. Report the latest terminal or non-terminal state and a support-ready request ID. | ||
@@ -44,0 +51,0 @@ |
| --- | ||
| name: mailrith-email-marketing | ||
| description: Plan and run safe Mailrith email marketing workflows through the Mailrith MCP server or CLI. Use for Mailrith subscriber sync, segmentation, campaign drafting, previews, approval-gated sends, automation changes, send-progress checks, activity review, permission diagnosis, and retry decisions. | ||
| description: Plan and run safe Mailrith email marketing workflows through the Mailrith MCP server or CLI. Use for Mailrith Subscriber sync, content, targeting, capture, Broadcast preparation and sending, Sequence operations, Automation changes, Webhooks, data transfer, progress checks, activity review, permission diagnosis, and retry decisions. | ||
| --- | ||
@@ -8,3 +8,3 @@ | ||
| Use Mailrith's discovery contract before choosing tools, request only the permissions the task needs, and keep people in control of high-impact work. | ||
| Use Mailrith's discovery contract before choosing tools and request only the permissions the task needs. | ||
@@ -16,8 +16,7 @@ ## Start Every Workflow | ||
| 3. Convert the user's request into the smallest sequence of Mailrith operations. | ||
| 4. Prefer reads and drafts first. Preview any operation that can send, publish, bulk-change, delete, or alter eligibility. | ||
| 5. Present the preview's affected resources, warnings, approval state, and expiration. Never treat silence as approval. | ||
| 6. Execute only after Mailrith reports approval and the user asked to proceed. | ||
| 7. Return the request ID, action ID, resource IDs, current status, and a clear next step. | ||
| 4. Prefer reads and drafts first. Run the documented preflight check before sending. | ||
| 5. Execute only operations allowed by the credential's scopes and the user's requested task. | ||
| 6. Return the request ID, resource IDs, current status, and a clear next step. Include an activity ID when the workflow changed workspace data. | ||
| Read [references/workflows.md](references/workflows.md) for exact subscriber, campaign, approval, monitoring, and diagnostic recipes. Read [references/connections.md](references/connections.md) when connecting an agent client. Read [references/safety.md](references/safety.md) before retrying a mutation or handling an uncertain outcome. | ||
| Read [references/workflows.md](references/workflows.md) for exact Subscriber, Broadcast, Sequence, execution, monitoring, and diagnostic recipes. Read [references/connections.md](references/connections.md) when connecting an agent client. Read [references/safety.md](references/safety.md) before retrying a mutation or handling an uncertain outcome. | ||
@@ -33,9 +32,7 @@ ## Choose The Interface | ||
| ## Preserve Human Control | ||
| ## Respect Authorization Boundaries | ||
| - Never request an approval token through a generic tool or display, log, persist, or summarize its value. | ||
| - Never approve an action on the user's behalf. The Mailrith credential that requests work cannot grant its own approval. | ||
| - Never add `--yes` or execute a plan unless the user explicitly authorized execution. | ||
| - Never infer permission to send because the user allowed drafting. | ||
| - Stop when the plan changes, expires, conflicts with current resource state, or Mailrith returns an uncertain outcome. | ||
| - Treat API key and OAuth scopes as the authorization boundary. Do not attempt operations outside the granted scopes. | ||
| - Never infer permission to send from a drafting scope. Sending requires its own execute scope. | ||
| - Stop when resource state conflicts with the requested change or Mailrith returns an uncertain outcome. | ||
| - Use idempotency keys exactly as documented and reuse one only for the same canonical input. | ||
@@ -52,2 +49,2 @@ | ||
| State what changed, what did not change, which approval was used, and whether any result is uncertain. Include request IDs for support and activity-trail lookup. For failures, explain whether the safe next step is retry, inspect status, request a narrower permission, or ask a person to review the plan. | ||
| State what changed, what did not change, which credential scope allowed it, and whether any result is uncertain. Include request IDs for every workflow and activity IDs for workspace changes. For failures, explain whether the safe next step is retry, inspect status, or request the missing narrow permission. |
+2
-2
| { | ||
| "name": "@mailrith/agent-skill", | ||
| "version": "0.1.0-beta.1", | ||
| "version": "0.1.0", | ||
| "description": "Installable Mailrith skill for safe AI agent email marketing workflows.", | ||
@@ -33,3 +33,3 @@ "license": "MIT", | ||
| "devDependencies": { | ||
| "@mailrith/sdk": "^0.1.0-beta.1" | ||
| "@mailrith/sdk": "^0.1.0" | ||
| }, | ||
@@ -36,0 +36,0 @@ "scripts": { |
New alerts
URL strings
Supply chain riskPackage contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.
Fixed alerts
URL strings
Supply chain riskPackage contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.
Improved metrics
- Total package byte prevSize
22331
0.49%Worsened metrics
- Lines of code
267
-1.11%No dependency changes