{"id":104064,"plugin_id":"plugin_asdk_app_6a6d31df2f848191acf2daa68c6ab8fb","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-11-18T18:11:13.836Z","digest":"cb3c8e34e631d880306162867c144a8b262e2fa9dc19593b9a31d7feb6ea472a","against":null,"payload":{"description":"Run a Sitelemetry audit on a website the user owns or is authorized to test and present the findings ranked by severity with the top fixes. Use when the user asks to audit, scan, check or score their own site for security, technical SEO, AI visibility, analytics and tracking, accessibility (WCAG) or performance (Core Web Vitals), asks for an overall site health check, or asks what an audit result, access message or allowance message means.","included_files":[{"relative_path":"agents/openai.yaml","size_in_bytes":342},{"relative_path":"assets/icon-large.png","size_in_bytes":33181},{"relative_path":"assets/icon-small.png","size_in_bytes":7770},{"relative_path":"references/result-handling.md","size_in_bytes":10236}],"name":"website-audit","skill_md_contents":"---\nname: website-audit\ndescription: Run a Sitelemetry audit on a website the user owns or is authorized to test and present the findings ranked by severity with the top fixes. Use when the user asks to audit, scan, check or score their own site for security, technical SEO, AI visibility, analytics and tracking, accessibility (WCAG) or performance (Core Web Vitals), asks for an overall site health check, or asks what an audit result, access message or allowance message means.\n---\n\n# Website audit\n\nUse the Sitelemetry audit tools to measure a website and explain what to fix first.\n\n## Ground rules\n\n- Follow the user's explicit instructions over the defaults in this skill. These rules always apply, even when the user asks otherwise: audit only sites the user owns or is authorized to test; never ask for passwords, API tokens or other credentials; never recommend a different account level, mention costs or suggest changing the account; never invent scores, findings or other results.\n- Only audit sites the user owns or is authorized to test. If that is not clear, ask before calling any tool. Do not audit competitors or other third-party sites, even when a tool would accept the target.\n- Reply in the user's language. Set `lang` to it when it is one of `en`, `tr`, `es`, `de`, `fr`, `pt`, `it`, `ja` or `zh`; otherwise leave `lang` out and translate your summary yourself. `lang` changes the wording of the result, not what is measured.\n- The audits run from Sitelemetry's servers. A completed audit uses the connected account's audit allowance. A result that says \"Audit not started\" uses none, and a failed audit releases what it reserved. An identical call made while the first run is still in progress continues that run; for a short time after it finishes, it can return the same result. Any other repeated or changed call can run another audit and use more allowance.\n- Security and full audits send test requests to the site. Depending on the account and on ownership verification, these can include active checks such as port and injection probes. Built-in checks send read-style requests and benign probes that are not meant to change the site but can appear in its server logs or firewall alerts. Say this before the first security or full audit of a site.\n- The host may ask the user to confirm each audit call. If the user declines, do not call the tool and do not report it as a failed audit.\n- Never invent scores, grades, findings, measurements or links.\n\n## 1. Get the target and the area\n\n- If the user's message names a site and an area, use them. If only a site is given and the area is unclear, ask, or use `audit_full` when the user wants everything.\n- `target` is a public `https://` URL or a bare hostname such as `example.com`. Localhost, private addresses, credentials in the URL and `#fragment` are not accepted.\n- Ownership verification covers a hostname. `example.com` and `www.example.com` are different hosts, so use the host the user verified in Sitelemetry for security and full audits.\n- Keep the same `target` spelling, `lang` and other arguments on every retry and re-test.\n\n## 2. Pick the narrowest tool\n\n| The user asks about | Tool | Optional arguments besides `target` and `lang` |\n|---|---|---|\n| Security headers, TLS, DNS, email authentication, exposed files | `audit_security` | `modules`, only when the user names specific checks |\n| Crawlability, indexation, metadata, structured data, internal links | `audit_seo` | `maxPages` (1-30, default 8), `includeSearchConsole` (default `true`), `searchConsoleDays` (`7`, `28` or `90`, default `28`) |\n| AI visibility, GEO, AEO, AI crawler policy | `audit_ai_visibility` | `maxPages` (1-30), `promptLimit` (1-24), `brand`, `competitors`; see the `ai-visibility` skill |\n| Analytics, tags, marketing pixels, consent, site verification tags | `audit_integrations` | none |\n| Accessibility, WCAG | `audit_accessibility` | none |\n| Speed, PageSpeed, Core Web Vitals | `audit_performance` | `strategy`: `mobile` (default) or `desktop` |\n| Everything, overall health, one combined score | `audit_full` | `strategy` for the performance area: `mobile` (default) or `desktop` |\n\n- Send only the arguments listed for the chosen tool. Anything else is rejected with \"Invalid audit arguments\". The optional arguments are for requests that need them; leave them out otherwise and the defaults apply.\n- `audit_security`: leave `modules` and `profile` out unless the user names checks; the default scope follows the connected account. `profile` accepts only `baseline`, which is already the default. Never send an empty `modules` list.\n- Module names for common requests: `dns`, `dns-email` (SPF, DKIM, DMARC), `rdap` (domain registration), `tls`, `http-headers`, `mitm-posture` (HTTPS enforcement), `tech-fingerprint` (server and framework disclosure), `http-methods` (HTTP methods and CORS), `exposure` (exposed files such as `.env`, `.git` or backups), `api-exposure` (OpenAPI, Swagger and GraphQL surface), `auth-surface` (admin and sign-in panels), `supply-chain` (exposed dependency manifests and lock files) and `wordpress`. The tool's input schema lists the rest.\n- `audit_seo` adds available Search Console evidence from the account's existing Google connection unless `includeSearchConsole` is `false`. This uses the account's Search Console keyword allowance, and missing access never stops the audit. Send `includeSearchConsole: false` when the user does not want Search Console data used, and `searchConsoleDays` only when the user asks for a 7-day or 90-day window.\n- `maxPages` on `audit_seo` and `audit_ai_visibility`: leave it out unless the user asks for a smaller or larger crawl. Default and explicit values are capped by the account's maximum and remaining allowance, so a result can cover fewer pages than requested; report what the result shows.\n- `audit_performance` sends the target URL to Google PageSpeed Insights for lab and field measurements.\n- `audit_accessibility` is a static WCAG 2.2 check, not a full manual accessibility review. Say so when it matters.\n- `audit_full` runs every audit area the connected account includes and returns a combined score across the measured areas. On an account with security-only access it runs security only. Its security area follows the `audit_security` defaults, and its performance area sends the target URL to Google PageSpeed Insights. Send `strategy: \"desktop\"` only when the user asks for desktop measurement.\n- Run one audit at a time and wait for each result before starting the next.\n\n## 3. Read the result\n\nCheck in this order. Every case is described in [result handling](references/result-handling.md); load it when a result is not a plain finished audit.\n\n1. **Error** (`isError`, text starting `Error:`): relay the message plainly and the correction it points to.\n2. **Prerequisite**: the text begins \"Audit not started. No audit quota was used.\", or `structuredContent.status` is `action_required`. No audit ran. Explain the `reason` (in `structuredContent` when present, otherwise from the text), pass on any link in the text unchanged, and do not retry until the prerequisite has changed. The empty score and findings are not results. For verification reasons, follow the `domain-verification` skill. If the `reason` is not one listed in [result handling](references/result-handling.md), for example a busy or temporarily unavailable service, no audit ran either: wait a little and retry the same call once.\n3. **Finished audit**: present it (section 4).\n\nAudits can take a few minutes. If a call times out or the connection drops, call the same tool once more with exactly the same arguments: while the first run is still in progress the service continues it, and for a short time after it finishes it can return the same result. Any other repeat can run another audit and use more allowance. Do not change arguments or start another audit meanwhile. If a result says `status: \"running\"` and includes `pollArguments`, wait about `retryAfterMs`, call the same tool with `pollArguments` exactly as returned, and never present it as finished.\n\n## 4. Present the findings\n\n1. **Summary line**: target, audit type, and `score` and `grade` (or `blended` for `audit_full`) only when they are not null. A null score means the area was not measured; it is not zero. Call `blended` a combined score only when `pillars` has all six areas with a numeric score and `failedPillars` is empty; otherwise name the areas that have a score.\n2. **Coverage**: what was not measured and why. For single-area tools use `status` (`partial`), `coverageStatus` and `auditDetails` when present. For `audit_full` use the areas present in `pillars`, null area scores, `failedPillars`, `complete` and the scope lines in the result text. A finished run does not mean every check was measured, and unmeasured checks are not passes.\n3. **Findings by severity**: critical, high, medium, low, then info. For each give the title, the affected location (the `Affected location:` line at the start of `evidence`, when present), the evidence, the impact and the fix (`remediation`). For `audit_full`, also name the `pillar`. If `findingsTruncated` is true, say that `returnedFindings` of `total` findings were returned.\n4. **Top fixes**: the three to five changes that remove the most severe risk first, in order, each tied to the findings it resolves.\n5. **Passing checks**: one line with `passingChecks` or the main passing items.\n\nKeep each finding's `findingKey`, the tool and the exact arguments you used, so a later re-test can be compared like for like. Offer the `fix-and-retest` skill as the next step. Results are returned in the conversation; there is no report link.\n\n## Access and allowance messages\n\nWhen a result says that an audit, security module or check is not included in the connected account, or that the audit allowance is used up, repeat that fact neutrally and pass on the link from the result text unchanged. Do not recommend a different account level, mention costs, or suggest that the user change their account. Do not describe or compare what other account levels include; if the user asks, point to the link in the result text. If the result describes a security-only alternative for the same target, you may mention it as an option and say that it does not complete the audit that was requested.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}