agentleFS
awaikawaik/direct-mcp-ai-project/.claude/skills/export-ad-reports/SKILL.md

export-ad-reports

Skill20 starsChanged 3 days agoMITView on GitHub

Выгружать отчёты Яндекс Директа, Метрики, VK Ads и Авито Рекламы через LidFly MCP v3 в подключённые Google Sheets или Google Docs с проверкой таблиц и публичных креативов. Использовать по запросам выгрузить, сохранить, дополнить или обновить статистику в Google-файле.

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. About 0 are common, 10 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill as the handoff between a provider skill, LidFly MCP v3 read calls, and Google connector write actions.

    line 8
  2. Distinctive:

    Never export OAuth or refresh tokens, API keys, passwords, seller secrets, signed private URLs, authorization headers, cookies, or other …

    line 14
  3. Distinctive:

    Remove secret fields from the payload and report the omission without treating ordinary advertising data as forbidden.

    line 15
  4. Very specific:

    Use the applicable provider skill to resolve scope: $yandex-direct-campaign-builder, $yandex-metrika, $vk-ads-campaign-builder, or …

    line 20
  5. Distinctive:

    Write explicit headers and data rows, then reread the exact changed range.

    line 29
  6. Very specific:

    If the connector exposes a native in-cell image or CellImage write action, prefer it and verify the resulting image value.

    line 30
  7. Very specific:

    Use a multi-argument formula only when cell sizing cannot meet the request.

    line 33
  8. Very specific:

    If an image cell returns a syntax error, replace it with the single-argument =IMAGE("URL") and reread that cell.

    line 35
  9. Very specific:

    Never replace an intended image with HYPERLINK, plain link text, or a link to an uploaded Drive file.

    line 37
  10. Very specific:

    Do not count Markdown such as !creative, a plain URL, or image alt text as an inserted image.

    line 43
  11. Distinctive:

    If the connector exposes no write action, say that no write action is available and recommend enabling write access, reconnecting Google …

    line 48
  12. Distinctive:

    If OAuth or permission fails, preserve and report the exact connector error and leave the target unchanged.

    line 49
  13. Distinctive:

    Do not infer that advertising data or public creatives are prohibited.

    line 50
  14. Distinctive:

    Never claim success from a hypothetical call, a local draft, or data displayed only in chat.

    line 51
  15. Distinctive:

    Treat these as required outcome branches:

    line 55
  16. Distinctive:

    No write actions: recommend write enablement, OAuth reconnect, or permission review.

    line 62

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Export Ad Reports

Use this skill as the handoff between a provider skill, LidFly MCP v3 read calls, and Google connector write actions. Do not write the export to a Workspace unless the user separately asks for that.

Data Boundary

  • Export aggregated statistics, spend, impressions, clicks, conversions, campaign/group/ad IDs, ad copy, and public creative URLs.
  • Export a creative image only when its URL is public and does not grant private access.
  • Never export OAuth or refresh tokens, API keys, passwords, seller secrets, signed private URLs, authorization headers, cookies, or other credentials.
  • Remove secret fields from the payload and report the omission without treating ordinary advertising data as forbidden.

Workflow

  1. Verify that the user explicitly requested a Google export or update. Resolve one exact target file and, for Sheets, the sheet and destination range or append position. If name lookup returns multiple files, list the candidates and request the file ID or URL; do not write while the target is ambiguous.
  2. Use the applicable provider skill to resolve scope: $yandex-direct-campaign-builder, $yandex-metrika, $vk-ads-campaign-builder, or $avito-ads. Use get_provider_context or resolve_campaign_scope when that skill requires it. A Google export alone does not require workspace_project_id.
  3. Find each report tool with search_tools, read its schema with get_tool_schema, and fetch the report with call_tool. Do not use call_write_tool merely to read or export a report.
  4. Restrict the export payload to the allowed data boundary and preserve the report period, currency, attribution, provider IDs, and aggregation level needed to interpret the numbers.
  5. Perform a real write with the connected Google Sheets, Google Docs, or Google Drive connector. External Google-file writes are connector actions, not LidFly MCP writes; never route them through call_write_tool.
  6. Reread the changed range or document through the Google connector and verify the written values and images. A prepared table, Markdown response, or successful provider read is not a completed export.
  7. Report the actual target file, changed sheet/range or document section, row/item count, report period and provider account or campaigns, verification result, and exact connector error if any. Mention Workspace (Пространство) only when the user asked to save there or a Workspace write actually happened; otherwise do not mention it or workspace_project_id at all. Never claim the export was saved to a project unless a Workspace write tool returned success.

Google Sheets

  • Write explicit headers and data rows, then reread the exact changed range.
  • If the connector exposes a native in-cell image or CellImage write action, prefer it and verify the resulting image value. The ordinary Google Sheets REST value/batch-update surface usually exposes formulas but not native CellImage creation.
  • Otherwise insert a public creative with the single-argument formula =IMAGE("URL"), using the original public URL from the provider response. Do not upload the creative to Drive for this formula: IMAGE does not support URLs hosted at drive.google.com.
  • Adjust image display size through row height and column width.
  • Use a multi-argument formula only when cell sizing cannot meet the request. For a Russian-locale sheet use semicolons, for example =IMAGE("URL";4;120;120).
  • Inspect reread values for #ERROR!, #N/A, #REF!, #VALUE!, and any other formula error.
  • If an image cell returns a syntax error, replace it with the single-argument =IMAGE("URL") and reread that cell.
  • If readback returns #REF!, another external-data access error, or the sheet displays an access banner, leave the =IMAGE(...) formula in the target cell. Ask an editor to open the spreadsheet in a desktop browser and click Allow access / «Разрешить доступ» once for that spreadsheet. After the user confirms, reread the image cells and continue verification.
  • Never replace an intended image with HYPERLINK, plain link text, or a link to an uploaded Drive file. A clickable link is not an image in the cell; if access remains unconfirmed, report the exact affected cells as pending instead of claiming a completed export.
  • Declare a table successful only after the expected rows are present and the changed range contains no unresolved formula errors.

Google Docs

  • Insert creatives as real inline images with a Google Docs connector action.
  • Do not count Markdown such as ![creative](URL), a plain URL, or image alt text as an inserted image.
  • Reread the document and verify the inline image/object in the document structure or in the connector's authoritative write result. If image insertion cannot be verified, report that limitation and do not claim complete success.

Connector Failures

  • If the connector exposes no write action, say that no write action is available and recommend enabling write access, reconnecting Google OAuth, or checking file permissions as appropriate.
  • If OAuth or permission fails, preserve and report the exact connector error and leave the target unchanged; do not silently create another file.
  • Do not infer that advertising data or public creatives are prohibited. Once the target is exact and a write action is available, make the real connector call before reporting a policy or connector block.
  • Never claim success from a hypothetical call, a local draft, or data displayed only in chat.

Completion Checks

Treat these as required outcome branches:

  • Successful Sheet: write, reread the target range, verify row count and zero formula errors.
  • Docs image: insert and verify a real inline image.
  • Russian locale: use ; only for a necessary multi-argument IMAGE formula.
  • Image syntax error: downgrade to single-argument IMAGE and reread.
  • External-data access prompt: leave IMAGE in place, request one-time «Разрешить доступ», and resume verification after confirmation; never fall back to a link.
  • No write actions: recommend write enablement, OAuth reconnect, or permission review.
  • OAuth/permission error: return the exact connector error without claiming success.
  • Ambiguous target: stop before writing and request the exact file ID or URL.

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.