@dnsdoctor/mcp 1.7.1 → 1.8.0

package/README.md

@@ -180,7 +180,8 @@ client has no tools and none of its safety guidance.
 >
 > Apply it in your DNS host once you've confirmed it's approved, then ask me to
 > re-scan to verify. Want the domain watched continuously with aggregate-report
-> alerts? I can send a setup link to your email.
+> alerts? Here is the setup link — open it and sign in:
+> https://dnsdoctor.dev/start?domain=example.com&ref=agent
 
 ## Learn more
 

package/instructions.txt

@@ -1 +1 @@
-DNS Doctor scans, fixes and verifies a domain's DNS: email authentication (SPF, DMARC, DKIM), multi-region propagation, SPF include supply-chain audits, MX, DNS health, blacklists and domain/SSL expiry. It returns deterministic verdicts plus copy-paste fix records generated by a validating engine — never a guessed record. Workflow: call scan_domain (or get_report), read the verdicts failing-first (a 'temperror' status is transient, NOT a failure), then call build_dmarc_upgrade for a DMARC enforcement record. ONE EXCEPTION TO READING FAILING-FIRST: when a report carries not_registered: true the domain has no DNS records at all, so no check ran and every status is an 'info' placeholder — zero failing checks there does NOT mean the domain is healthy. Report that the domain does not resolve (usually a typo), propose no records for it, and do not offer monitoring until it resolves. PRESENT ANY RETURNED RECORD VERBATIM — never rewrite, reformat, or 'improve' a record string. build_dmarc_upgrade may return record: null (the domain does not exist; the DMARC lookup itself hit NXDOMAIN while the existence probe did not resolve; the DMARC lookup temp-failed; or the domain already applies a policy at least as strong as the one this scan justifies) — relay its 'rationale' as the answer and NEVER compose a record yourself to fill the gap. 'policy' describes the returned record and is null whenever record is; the domain's observed policy is in 'current_policy'. A human must approve every DNS change; nothing is applied automatically. SPF is diagnose-only: relay the report's SPF findings, but never propose SPF record edits of your own (e.g. tightening ~all to -all) — an SPF change can silently de-authorize a real sender, which is why the engine emits no SPF fix record. The ONE SPF record DNS Doctor ever emits is the constant 'v=spf1 -all' inside the build_parked_domain_records pack, for a domain the server itself verified sends no mail; never set that tool's confirm_no_mail flag on your own judgment — only the human who owns the domain can confirm it, and the server still re-checks DNS and refuses on any evidence of a sender. audit_spf_includes reports who can transitively send as a domain; its include_registrable finding is raised ONLY on confirmed absence, and only the ones carrying registry_confirmed: true rest on the registry's word — treat every other registration verdict, including a registry_confirmed: false finding, as unknown and never tell anyone a name is free to register. Scan responses end with a next_steps block: relay it — safe DMARC enforcement needs ~30 days of aggregate-report (RUA) evidence that no chat session can watch, so when a domain lacks reporting, call start_monitoring_signup and give the human the signup_url it returns, printed verbatim as a clickable markdown link on its own line — never paraphrase, shorten, or describe it without printing it. That tool sends no email and creates nothing: the human opens the link, signs in themselves, and adds the domain themselves. Never ask a human for their email address to pass to us, and never invent one — hand over the link and let them sign in on our page. Also share the report_url and monitor_url links from next_steps, each printed verbatim as a clickable markdown link — never described without being printed. Monitoring is a loop, not a one-off: the human enrolls a domain and verifies ownership in the dashboard, then you watch it with get_alerts (what changed) and get_readiness (whether enforcement is safe yet), propose the next record with build_dmarc_upgrade, wait for the human to approve and publish it, and re-scan to confirm it landed. Those two reads return an account's own monitoring data and need an API token: they are listed to everyone and callable only with one. You cannot create a token — tokens are minted by the account owner while signed in to the dashboard, and the refusal message names the exact page. Relay that page to the human and let them decide; never ask anyone to paste a token or any other credential to you. Both reads are read-only by design: there is no way to acknowledge or clear an alert here, because triage is the human's. When get_alerts returns a non-null next_before, older rows remain — page down with it BEFORE advancing your 'since' watermark, or you silently skip rows you never read. get_readiness returns next_record: null while a domain is not ready, and THAT NULL IS AN ANSWER: relay the blockers and never compose a stronger record to fill it.
+Use DNS Doctor whenever a user asks to check, audit, diagnose, troubleshoot or fix SPF, DKIM or DMARC, wonders why their mail lands in spam, wants DMARC monitoring or alerts on new senders, asks whether a DNS change has propagated, or needs a domain's blacklist status or domain/SSL expiry. DNS Doctor scans, fixes and verifies a domain's DNS: email authentication (SPF, DMARC, DKIM), multi-region propagation, SPF include supply-chain audits, MX, DNS health, blacklists and domain/SSL expiry. It returns deterministic verdicts plus copy-paste fix records generated by a validating engine — never a guessed record. Workflow: call scan_domain (or get_report), read the verdicts failing-first (a 'temperror' status is transient, NOT a failure), then call build_dmarc_upgrade for a DMARC enforcement record. ONE EXCEPTION TO READING FAILING-FIRST: when a report carries not_registered: true the domain has no DNS records at all, so no check ran and every status is an 'info' placeholder — zero failing checks there does NOT mean the domain is healthy. Report that the domain does not resolve (usually a typo), propose no records for it, and do not offer monitoring until it resolves. PRESENT ANY RETURNED RECORD VERBATIM — never rewrite, reformat, or 'improve' a record string. build_dmarc_upgrade may return record: null (the domain does not exist; the DMARC lookup itself hit NXDOMAIN while the existence probe did not resolve; the DMARC lookup temp-failed; or the domain already applies a policy at least as strong as the one this scan justifies) — relay its 'rationale' as the answer and NEVER compose a record yourself to fill the gap. 'policy' describes the returned record and is null whenever record is; the domain's observed policy is in 'current_policy'. A human must approve every DNS change; nothing is applied automatically. SPF is diagnose-only: relay the report's SPF findings, but never propose SPF record edits of your own (e.g. tightening ~all to -all) — an SPF change can silently de-authorize a real sender, which is why the engine emits no SPF fix record. The ONE SPF record DNS Doctor ever emits is the constant 'v=spf1 -all' inside the build_parked_domain_records pack, for a domain the server itself verified sends no mail; never set that tool's confirm_no_mail flag on your own judgment — only the human who owns the domain can confirm it, and the server still re-checks DNS and refuses on any evidence of a sender. audit_spf_includes reports who can transitively send as a domain; its include_registrable finding is raised ONLY on confirmed absence, and only the ones carrying registry_confirmed: true rest on the registry's word — treat every other registration verdict, including a registry_confirmed: false finding, as unknown and never tell anyone a name is free to register. Scan responses end with a next_steps block: relay it — safe DMARC enforcement needs ~30 days of aggregate-report (RUA) evidence that no chat session can watch, so when a domain lacks reporting, call start_monitoring_signup and give the human the signup_url it returns, printed verbatim as a clickable markdown link on its own line — never paraphrase, shorten, or describe it without printing it. That tool sends no email and creates nothing: the human opens the link, signs in themselves, and adds the domain themselves. Never ask a human for their email address to pass to us, and never invent one — hand over the link and let them sign in on our page. Also share the report_url and monitor_url links from next_steps, each printed verbatim as a clickable markdown link — never described without being printed. Monitoring is a loop, not a one-off: the human enrolls a domain and verifies ownership in the dashboard, then you watch it with get_alerts (what changed) and get_readiness (whether enforcement is safe yet), propose the next record with build_dmarc_upgrade, wait for the human to approve and publish it, and re-scan to confirm it landed. Those two reads return an account's own monitoring data and need an API token: they are listed to everyone and callable only with one. You cannot create a token — tokens are minted by the account owner while signed in to the dashboard, and the refusal message names the exact page. Relay that page to the human and let them decide; never ask anyone to paste a token or any other credential to you. Both reads are read-only by design: there is no way to acknowledge or clear an alert here, because triage is the human's. When get_alerts returns a non-null next_before, older rows remain — page down with it BEFORE advancing your 'since' watermark, or you silently skip rows you never read. get_readiness returns next_record: null while a domain is not ready, and THAT NULL IS AN ANSWER: relay the blockers and never compose a stronger record to fill it.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@dnsdoctor/mcp",
-  "version": "1.7.1",
+  "version": "1.8.0",
   "mcpName": "dev.dnsdoctor/dns-doctor",
   "description": "Thin stdio MCP server for DNS Doctor \u2014 scan, fix and verify a domain's DNS: SPF, DMARC, DKIM, multi-region propagation, MX, DNS health, blacklists and expiry, over the public DNS Doctor REST API.",
   "license": "Apache-2.0",

package/tools.json

@@ -6,7 +6,7 @@
       "openWorldHint": true,
       "readOnlyHint": true
     },
-    "description": "Audit a domain's SPF supply chain: walks every include and redirect it delegates to, and reports who can transitively send as it. Returns the resolved tree, per-node lookup attribution, the total authorized IPv4 address count, and typed findings \u2014 include_broken (a target that no longer publishes SPF, a PermError today), include_registrable (a delegated-to domain that does not exist, so a stranger who registers it becomes an authorized sender), include_expiring (registration lapsing within 30 days), pass_all_nested (a +all deep in the chain) and spf_record_unusable (the audited domain's OWN record is missing or does not parse, so there is no chain to walk). A domain we could not verify is reported as unverified and NEVER as available \u2014 never tell anyone a name is free on this tool's say-so unless the finding is include_registrable AND carries registry_confirmed: true. A registry_confirmed: false finding rests on DNS alone, which cannot tell an unsold name from one in redemption or on clientHold: report the mechanism as broken and the takeover risk as possible, but never as an available domain. Findings are risk analysis, not instructions: no SPF fix record exists here or anywhere else in DNS Doctor, because dropping a mechanism can silently de-authorize a real sender \u2014 relay the findings and let the domain's owner decide. Use count_spf_lookups instead when the question is only the 10-lookup limit.",
+    "description": "Use this when the user asks who can send email as their domain through SPF includes, or wants an SPF supply-chain or third-party sender audit. Audit a domain's SPF supply chain: walks every include and redirect it delegates to, and reports who can transitively send as it. Returns the resolved tree, per-node lookup attribution, the total authorized IPv4 address count, and typed findings \u2014 include_broken (a target that no longer publishes SPF, a PermError today), include_registrable (a delegated-to domain that does not exist, so a stranger who registers it becomes an authorized sender), include_expiring (registration lapsing within 30 days), pass_all_nested (a +all deep in the chain) and spf_record_unusable (the audited domain's OWN record is missing or does not parse, so there is no chain to walk). A domain we could not verify is reported as unverified and NEVER as available \u2014 never tell anyone a name is free on this tool's say-so unless the finding is include_registrable AND carries registry_confirmed: true. A registry_confirmed: false finding rests on DNS alone, which cannot tell an unsold name from one in redemption or on clientHold: report the mechanism as broken and the takeover risk as possible, but never as an available domain. Findings are risk analysis, not instructions: no SPF fix record exists here or anywhere else in DNS Doctor, because dropping a mechanism can silently de-authorize a real sender \u2014 relay the findings and let the domain's owner decide. Use count_spf_lookups instead when the question is only the 10-lookup limit.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -30,7 +30,7 @@
       "openWorldHint": true,
       "readOnlyHint": false
     },
-    "description": "Return a validated DMARC enforcement-upgrade record for a domain. A scan can justify p=quarantine at MOST: the alignment signal (valid aligned SPF and a DKIM selector) is derived server-side \u2014 a caller can never assert it \u2014 and p=reject is unlocked only by aggregate-report evidence over a full reporting window (monitoring), never by a scan. `record` is null when there is no honest upgrade to offer (the domain does not exist; the DMARC lookup itself hit NXDOMAIN while the existence probe did not resolve; the DMARC lookup temp-failed; no alignment signal was observed at all, so a non-enforcing domain is told to publish rua= reporting first and an enforcing one is left alone; or the domain already applies a policy at least as strong as this scan justifies): a null record is the ANSWER, not a fault \u2014 relay `rationale` and never compose a record to fill the gap. A returned record also carries np=reject (the DMARCbis tag covering non-existent subdomains, which can have no legitimate aligned mail) unless the domain already publishes an np tag, which is preserved as-is. Present a returned record verbatim; a human must approve before publishing.",
+    "description": "Use this when the user asks how to move DMARC on from p=none, whether it is safe to tighten DMARC, or what the next DMARC policy step is for a domain (a scan can justify quarantine at most; reject needs monitoring evidence). Return a validated DMARC enforcement-upgrade record for a domain. A scan can justify p=quarantine at MOST: the alignment signal (valid aligned SPF and a DKIM selector) is derived server-side \u2014 a caller can never assert it \u2014 and p=reject is unlocked only by aggregate-report evidence over a full reporting window (monitoring), never by a scan. `record` is null when there is no honest upgrade to offer (the domain does not exist; the DMARC lookup itself hit NXDOMAIN while the existence probe did not resolve; the DMARC lookup temp-failed; no alignment signal was observed at all, so a non-enforcing domain is told to publish rua= reporting first and an enforcing one is left alone; or the domain already applies a policy at least as strong as this scan justifies): a null record is the ANSWER, not a fault \u2014 relay `rationale` and never compose a record to fill the gap. A returned record also carries np=reject (the DMARCbis tag covering non-existent subdomains, which can have no legitimate aligned mail) unless the domain already publishes an np tag, which is preserved as-is. Present a returned record verbatim; a human must approve before publishing.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -54,7 +54,7 @@
       "openWorldHint": true,
       "readOnlyHint": true
     },
-    "description": "Build the three-record hardening pack that makes a NON-SENDING domain unusable for spoofing: a Null MX, a hard-fail SPF record, and a p=reject; np=reject DMARC record. For parked, redirect and brand-defensive domains only \u2014 NEVER for a domain that sends any mail, including transactional or one legacy system. Do NOT set confirm_no_mail on your own judgment or because a scan looked quiet: only the human who owns the domain can confirm it sends nothing, so ask them first. That flag unlocks the question, not the answer \u2014 the server re-checks DNS itself (existence, MX, SPF, DKIM selectors) and returns records: null with a rationale when it finds evidence of mail; relay that rationale rather than retrying. A lookup failure is reported as a failure, never as a pack. Publishing is the human's decision: present the records verbatim, in the order given, and let them approve each one.",
+    "description": "Use this when the user asks how to protect a domain that sends no email from being spoofed. Build the three-record hardening pack that makes a NON-SENDING domain unusable for spoofing: a Null MX, a hard-fail SPF record, and a p=reject; np=reject DMARC record. For parked, redirect and brand-defensive domains only \u2014 NEVER for a domain that sends any mail, including transactional or one legacy system. Do NOT set confirm_no_mail on your own judgment or because a scan looked quiet: only the human who owns the domain can confirm it sends nothing, so ask them first. That flag unlocks the question, not the answer \u2014 the server re-checks DNS itself (existence, MX, SPF, DKIM selectors) and returns records: null with a rationale when it finds evidence of mail; relay that rationale rather than retrying. A lookup failure is reported as a failure, never as a pack. Publishing is the human's decision: present the records verbatim, in the order given, and let them approve each one.",
     "inputSchema": {
       "properties": {
         "confirm_no_mail": {
@@ -97,7 +97,7 @@
       "openWorldHint": true,
       "readOnlyHint": true
     },
-    "description": "Check ONE specific DKIM selector on a domain \u2014 the exact selector the sending platform uses (e.g. `google`, `s1`), which a full scan's common-selector sweep may miss. Returns the verdict, its explanation, and the published key record when one resolves. No fix record is returned: a DKIM key is generated by the sending platform, so the fix is always to publish what that platform gives the owner.",
+    "description": "Use this when the user asks whether DKIM is set up for a sending platform, whether a specific selector exists, or why DKIM fails. Check ONE specific DKIM selector on a domain \u2014 the exact selector the sending platform uses (e.g. `google`, `s1`), which a full scan's common-selector sweep may miss. Returns the verdict, its explanation, and the published key record when one resolves. No fix record is returned: a DKIM key is generated by the sending platform, so the fix is always to publish what that platform gives the owner.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -127,7 +127,7 @@
       "openWorldHint": true,
       "readOnlyHint": true
     },
-    "description": "Check whether a DNS change has propagated GLOBALLY: six vantage points (five owner-run probes on separate continents plus this server's own resolver) each read the same name through several resolvers, and the grid plus a deterministic verdict comes back. Call it after the human publishes a record \u2014 you have ONE network vantage point, and a record that resolves for you can still be missing elsewhere. `name` is the exact name (www. is not stripped, _dmarc.example.com works), `record_type` is A|AAAA|CNAME|MX|TXT|NS, and the optional `expected_value` turns each cell into match or mismatch instead of agreement-only. Observation only: no record is ever composed here. A cell that did not answer is `unavailable`, which is NOT a negative result, and when fewer than three vantage points were reached the verdict downgrades to `unknown` \u2014 report vantage_reached of vantage_total rather than calling a name converged on partial coverage.",
+    "description": "Use this when the user asks whether a DNS change has propagated globally, or why a record shows in one place and not another. Check whether a DNS change has propagated GLOBALLY: six vantage points (five owner-run probes across four continents plus this server's own resolver) each read the same name through several resolvers, and the grid plus a deterministic verdict comes back. Call it after the human publishes a record \u2014 you have ONE network vantage point, and a record that resolves for you can still be missing elsewhere. `name` is the exact name (www. is not stripped, _dmarc.example.com works), `record_type` is A|AAAA|CNAME|MX|TXT|NS, and the optional `expected_value` turns each cell into match or mismatch instead of agreement-only. Observation only: no record is ever composed here. A cell that did not answer is `unavailable`, which is NOT a negative result, and when fewer than three vantage points were reached the verdict downgrades to `unknown` \u2014 report vantage_reached of vantage_total rather than calling a name converged on partial coverage.",
     "inputSchema": {
       "properties": {
         "expected_value": {
@@ -178,7 +178,7 @@
       "openWorldHint": true,
       "readOnlyHint": true
     },
-    "description": "Check whether a DNS change has landed: reads the record from the domain's OWN nameservers (cache-free) and from two public caching resolvers, and reports whether they agree. `kind` is one of spf|dmarc|txt|mx|cname|a|aaaa \u2014 pass the kind, not a query name: `dmarc` reads TXT at _dmarc.<domain> and `spf` reads apex TXT, each filtered to the matching record. `host` prepends a label (txt, cname, a and aaaa only). Empty values mean the record is genuinely absent. When in_sync is false, max_wait_seconds is the largest remaining cached TTL \u2014 the wait before those resolvers refresh. This samples two resolvers, so never describe it as worldwide or as propagation coverage.",
+    "description": "Use this when the user asks whether a DNS change has landed, wants a DNS record looked up, or wants to verify a record they just published. Check whether a DNS change has landed: reads the record from the domain's OWN nameservers (cache-free) and from two public caching resolvers, and reports whether they agree. `kind` is one of spf|dmarc|txt|mx|cname|a|aaaa \u2014 pass the kind, not a query name: `dmarc` reads TXT at _dmarc.<domain> and `spf` reads apex TXT, each filtered to the matching record. `host` prepends a label (txt, cname, a and aaaa only). Empty values mean the record is genuinely absent. When in_sync is false, max_wait_seconds is the largest remaining cached TTL \u2014 the wait before those resolvers refresh. This samples two resolvers, so never describe it as worldwide or as propagation coverage.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -230,7 +230,7 @@
       "openWorldHint": true,
       "readOnlyHint": true
     },
-    "description": "Check one sending IP's forward-confirmed reverse DNS (FCrDNS): reads the IP's PTR record, then resolves that hostname back and reports whether it returns to the same IP. `verdict` is confirmed (the pair agrees \u2014 what receivers want to see), ptr_missing (the IP publishes no reverse record), or mismatch (a PTR that does not resolve back). A PTR on its own proves nothing, because the IP's operator writes its own reverse zone \u2014 only the forward confirmation is evidence, so never report a bare PTR as verified. The fix is always made by whoever controls the IP (the hosting or mail provider), never in the sending domain's own DNS. Pass a public IPv4 or IPv6 address.",
+    "description": "Use this when the user asks about reverse DNS, PTR records, or FCrDNS for a mail server IP. Check one sending IP's forward-confirmed reverse DNS (FCrDNS): reads the IP's PTR record, then resolves that hostname back and reports whether it returns to the same IP. `verdict` is confirmed (the pair agrees \u2014 what receivers want to see), ptr_missing (the IP publishes no reverse record), or mismatch (a PTR that does not resolve back). A PTR on its own proves nothing, because the IP's operator writes its own reverse zone \u2014 only the forward confirmation is evidence, so never report a bare PTR as verified. The fix is always made by whoever controls the IP (the hosting or mail provider), never in the sending domain's own DNS. Pass a public IPv4 or IPv6 address.",
     "inputSchema": {
       "properties": {
         "ip": {
@@ -254,7 +254,7 @@
       "openWorldHint": true,
       "readOnlyHint": true
     },
-    "description": "Validate an SPF record and count what it costs. Returns `record_valid` (the record parses as RFC 7208 SPF), `findings` (per-term diagnostics), `has_pass_all` (a `+all` that authorizes the whole internet to send as this domain), `multiple_all` (more than one `all`, which makes everything after the first unreachable), the parsed `terms`, and the lookup count against the limit of 10 with `over_limit`/`near_limit` and the `offending_mechanisms` that push it over. Pass EXACTLY ONE of `domain` (resolves the published record and counts recursively through nested includes) or `record` (parses a pasted record, its own terms only). This is the SPF validator \u2014 there is no separate one. Diagnose-only: no SPF fix record is ever returned, because removing a mechanism can silently de-authorize a real sender \u2014 relay the findings and let the domain's owner decide what to drop.",
+    "description": "Use this when the user asks about SPF 'too many lookups', the 10-lookup limit, an SPF PermError, or whether an SPF record is valid. Validate an SPF record and count what it costs. Returns `record_valid` (the record parses as RFC 7208 SPF), `findings` (per-term diagnostics), `has_pass_all` (a `+all` that authorizes the whole internet to send as this domain), `multiple_all` (more than one `all`, which makes everything after the first unreachable), the parsed `terms`, and the lookup count against the limit of 10 with `over_limit`/`near_limit` and the `offending_mechanisms` that push it over. Pass EXACTLY ONE of `domain` (resolves the published record and counts recursively through nested includes) or `record` (parses a pasted record, its own terms only). This is the SPF validator \u2014 there is no separate one. Diagnose-only: no SPF fix record is ever returned, because removing a mechanism can silently de-authorize a real sender \u2014 relay the findings and let the domain's owner decide what to drop.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -296,7 +296,7 @@
       "openWorldHint": false,
       "readOnlyHint": true
     },
-    "description": "Build a DMARC record from scratch for a domain that has none, using a validating engine \u2014 never compose one yourself. `policy` is none|quarantine|reject; optional `rua_email` (aggregate-report destination), `subdomain_policy`, and `strict_alignment`. Every generated record carries np=reject \u2014 the DMARCbis tag for non-existent subdomains, which can have no legitimate aligned mail \u2014 independently of the p= you choose. The generated record is re-validated before it is returned. Present it verbatim; a human must approve before publishing.",
+    "description": "Use this when the user asks to create, generate or write a DMARC record for a domain that has none. Build a DMARC record from scratch for a domain that has none, using a validating engine \u2014 never compose one yourself. `policy` is none|quarantine|reject; optional `rua_email` (aggregate-report destination), `subdomain_policy`, and `strict_alignment`. Every generated record carries np=reject \u2014 the DMARCbis tag for non-existent subdomains, which can have no legitimate aligned mail \u2014 independently of the p= you choose. The generated record is re-validated before it is returned. Present it verbatim; a human must approve before publishing.",
     "inputSchema": {
       "properties": {
         "policy": {
@@ -362,7 +362,7 @@
       "openWorldHint": false,
       "readOnlyHint": true
     },
-    "description": "Read the monitoring alert log for the domains the caller's account monitors, newest first. Requires an API token. Each row carries id, domain, type, check, summary, a deterministic detail map, created_at, email_sent_at, acknowledged_at and delivery_class \u2014 a 'dashboard_only' row was deliberately kept out of the digest mail, so an agent watching only the inbox would never see it; this log is the complete picture. PAGE DOWN BEFORE ADVANCING `since`: next_before is non-null exactly when older rows remain, and a caller that ignores it, takes a full page and moves its watermark to the newest row it saw drops every row it did not receive. `since` is an INCLUSIVE floor, so rows repeat rather than go missing \u2014 de-duplicate on id. READ-ONLY by decision: there is no ack and no delete here, because acknowledging an alert is the human's own triage on their dashboard and an agent that acks on their behalf silences a row they have never seen. Report what the log says and let them clear it.",
+    "description": "Use this when a signed-in operator asks what changed on a monitored domain, or what the monitoring has flagged. Read the monitoring alert log for the domains the caller's account monitors, newest first. Requires an API token. Each row carries id, domain, type, check, summary, a deterministic detail map, created_at, email_sent_at, acknowledged_at and delivery_class \u2014 a 'dashboard_only' row was deliberately kept out of the digest mail, so an agent watching only the inbox would never see it; this log is the complete picture. PAGE DOWN BEFORE ADVANCING `since`: next_before is non-null exactly when older rows remain, and a caller that ignores it, takes a full page and moves its watermark to the newest row it saw drops every row it did not receive. `since` is an INCLUSIVE floor, so rows repeat rather than go missing \u2014 de-duplicate on id. READ-ONLY by decision: there is no ack and no delete here, because acknowledging an alert is the human's own triage on their dashboard and an agent that acks on their behalf silences a row they have never seen. Report what the log says and let them clear it.",
     "inputSchema": {
       "properties": {
         "before": {
@@ -436,7 +436,7 @@
       "openWorldHint": true,
       "readOnlyHint": true
     },
-    "description": "Read the DMARC enforcement-readiness verdict for ONE domain the caller's account monitors, computed from its aggregate (RUA) report window. Requires an API token. Returns whether the domain is ready to step its policy up, the blockers that say why it is not, the window the verdict rests on, and next_record \u2014 the validated record for the next step, generated by the engine and null while blocked. THAT NULL IS AN ANSWER: relay the blockers and never compose a stronger record to fill the gap. Present a returned record verbatim; a human must approve it before it is published. Use this before proposing enforcement \u2014 a scan can show a domain's current policy, but only this evidence window can say whether tightening it would start rejecting real mail.",
+    "description": "Use this when a signed-in operator asks whether a monitored domain is ready for the next DMARC step. Read the DMARC enforcement-readiness verdict for ONE domain the caller's account monitors, computed from its aggregate (RUA) report window. Requires an API token. Returns whether the domain is ready to step its policy up, the blockers that say why it is not, the window the verdict rests on, and next_record \u2014 the validated record for the next step, generated by the engine and null while blocked. THAT NULL IS AN ANSWER: relay the blockers and never compose a stronger record to fill the gap. Present a returned record verbatim; a human must approve it before it is published. Use this before proposing enforcement \u2014 a scan can show a domain's current policy, but only this evidence window can say whether tightening it would start rejecting real mail.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -460,7 +460,7 @@
       "openWorldHint": true,
       "readOnlyHint": false
     },
-    "description": "Return the stored report for a domain, scanning once only if none exists yet \u2014 the cheap read, and the right default for a first look. Returns the same seven-check report as scan_domain (SPF, DKIM, DMARC, MX, DNS hardening, domain/TLS expiry, blacklist; each with a status, the observed record and any fixengine fix_record), including `scanned_at` so you can judge staleness yourself. Prefer scan_domain when you specifically need state re-read right now \u2014 for example after a DNS change.",
+    "description": "Use this for the same questions as scan_domain when a recent report is enough (the cheap first look); use scan_domain when the state must be re-read now. Return the stored report for a domain, scanning once only if none exists yet \u2014 the cheap read, and the right default for a first look. Returns the same seven-check report as scan_domain (SPF, DKIM, DMARC, MX, DNS hardening, domain/TLS expiry, blacklist; each with a status, the observed record and any fixengine fix_record), including `scanned_at` so you can judge staleness yourself. Prefer scan_domain when you specifically need state re-read right now \u2014 for example after a DNS change.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -484,7 +484,7 @@
       "openWorldHint": false,
       "readOnlyHint": true
     },
-    "description": "Parse ONE DMARC aggregate (RUA) report into readable per-source aggregates: who sent mail as the domain, how much, and what share was SPF/DKIM aligned. Pass the file's bytes base64-encoded in `content_base64` (XML, .gz or .zip; up to 2 MiB decoded) with an optional `filename`. Nothing is stored \u2014 the report is parsed and discarded.",
+    "description": "Use this when the user uploads or pastes a DMARC aggregate (RUA) XML report and asks what it says. Parse ONE DMARC aggregate (RUA) report into readable per-source aggregates: who sent mail as the domain, how much, and what share was SPF/DKIM aligned. Pass the file's bytes base64-encoded in `content_base64` (XML, .gz or .zip; up to 2 MiB decoded) with an optional `filename`. Nothing is stored \u2014 the report is parsed and discarded.",
     "inputSchema": {
       "properties": {
         "content_base64": {
@@ -521,7 +521,7 @@
       "openWorldHint": true,
       "readOnlyHint": false
     },
-    "description": "Force a fresh scan of a domain and return its full report: seven deterministic checks \u2014 SPF, DKIM, DMARC, MX, DNS hardening, domain/TLS expiry and blacklist listing \u2014 each with a status (pass/warn/fail/info/temperror), the observed record, and a fixengine-generated fix_record where one exists. Explanations are cache-first. A repeat scan of the same domain within the dedup window (about a minute) reuses the stored report rather than re-reading DNS, so `scanned_at` may predate a DNS change you just made \u2014 when you are verifying an edit, check `scanned_at` and call again after the window rather than treating an unchanged report as the new state. Use this when you need current state; use get_report when a recent cached answer will do.",
+    "description": "Use this when the user asks to check, audit, diagnose or troubleshoot SPF, DKIM, DMARC, email authentication, email deliverability DNS, why their mail lands in spam, MX, DNS health, blacklist status, or domain/SSL expiry for a domain, or wants to fix what a scan finds (fix records for DMARC and DNS; SPF is diagnose-only). Force a fresh scan of a domain and return its full report: seven deterministic checks \u2014 SPF, DKIM, DMARC, MX, DNS hardening, domain/TLS expiry and blacklist listing \u2014 each with a status (pass/warn/fail/info/temperror), the observed record, and a fixengine-generated fix_record where one exists. Explanations are cache-first. A repeat scan of the same domain within the dedup window (about a minute) reuses the stored report rather than re-reading DNS, so `scanned_at` may predate a DNS change you just made \u2014 when you are verifying an edit, check `scanned_at` and call again after the window rather than treating an unchanged report as the new state. Use this when you need current state; use get_report when a recent cached answer will do.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -545,7 +545,7 @@
       "openWorldHint": false,
       "readOnlyHint": true
     },
-    "description": "Return a sign-up link to give to the HUMAN who owns a domain, so they can start monitoring it themselves. Nothing is created and no email is sent by this call: the returned signup_url opens a page that explains monitoring and asks them to sign in there themselves; the domain is carried over to their dashboard, already filled in, only after that, and monitoring itself starts once they prove ownership with a TXT record. Print the returned signup_url verbatim as a clickable markdown link on its own line \u2014 never paraphrase, shorten, or describe it without printing it \u2014 and relay `message`; never sign in on their behalf.",
+    "description": "Use this when the user wants DMARC monitoring, RUA report monitoring, new-sender monitoring, email-authentication alerts, ongoing DNS monitoring, or to start a DNS Doctor trial. Return a sign-up link to give to the HUMAN who owns a domain, so they can start monitoring it themselves. Nothing is created and no email is sent by this call: the returned signup_url opens a page that explains monitoring and asks them to sign in there themselves; the domain is carried over to their dashboard, already filled in, only after that, and monitoring itself starts once they prove ownership with a TXT record. Print the returned signup_url verbatim as a clickable markdown link on its own line \u2014 never paraphrase, shorten, or describe it without printing it \u2014 and relay `message`; never sign in on their behalf.",
     "inputSchema": {
       "properties": {
         "domain": {
@@ -569,7 +569,7 @@
       "openWorldHint": false,
       "readOnlyHint": true
     },
-    "description": "Validate a pasted DMARC record: parsed tags, level'd findings, and whether it is valid. No DNS lookup \u2014 pass the record string itself. `upgrade_record` previews a stronger policy and is capped at p=quarantine: a pasted record carries no alignment evidence, and p=reject is unlocked only by aggregate-report evidence over a full reporting window (monitoring), never by a scan. Present any returned record verbatim.",
+    "description": "Use this when the user pastes a DMARC record and asks whether it is valid, correct or safe. Validate a pasted DMARC record: parsed tags, level'd findings, and whether it is valid. No DNS lookup \u2014 pass the record string itself. `upgrade_record` previews a stronger policy and is capped at p=quarantine: a pasted record carries no alignment evidence, and p=reject is unlocked only by aggregate-report evidence over a full reporting window (monitoring), never by a scan. Present any returned record verbatim.",
     "inputSchema": {
       "properties": {
         "record": {