add-connector
Skill11 starsChanged 4 days agoMITView on GitHub
Research-first evaluation of a third-party data source for Duct (OAuth, scopes, quotas, APIs) using official docs via web search, then map findings to repo structure—no step-by-step coding walkthrough.
5 other projects use a version of this.
Adopted by
Most recently taken up by
- Azure/connectors-python-sdktheir own version, first seen Oct 27, 2026
- pipeshub-ai/pipeshub-aitheir own version, first seen Oct 25, 2026
- microsoft/power-platform-skillstheir own version, first seen Oct 13, 2026
- Ame-Al/ametheir own version, first seen Oct 8, 2026
- simstudioai/simtheir own version, first seen Oct 8, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- simstudioai/sim★ 30ksame name
- pipeshub-ai/pipeshub-ai★ 4.1ksame name
- microsoft/power-platform-skills★ 1.2ksame name
- Ame-Al/ame★ 1same name
- Azure/connectors-python-sdk★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- app/AGENTS.mdAGENTS.md
- backend/AGENTS.mdAGENTS.md
- cli/AGENTS.mdAGENTS.md
- desktop/AGENTS.mdAGENTS.md
- site/AGENTS.mdAGENTS.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/landing-pages.mdcCursor rule
- site/llms.txtllms.txt
- add-blog-postSkill
- add-changelog-entrySkill
- component-previewSkill
- featureSkill
- launch-readiness-auditSkill
- mosaic-panelSkill
- new-pageSkill
- prioritizeSkill
- security-auditSkill
- session-auditSkill
- working-backwardsSkill
- duct-sprint-refreshSkill
- launch-checkSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 7 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
7
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 29
If the user does not ask for a file, deliver the Connector research memo in chat.
- Very specific: line 37
Start official: Search vendor developer docs (e.g.
- Distinctive: line 38
Follow up: Run additional searches for OAuth app verification / consent, quotas, pagination, API versioning, deprecation notices, and …
- Distinctive: line 40
Conflicts: If sources disagree, vendor documentation wins over Stack Overflow or generic tutorials.
- Distinctive: line 41
Deep pages: When search results are vague, open or fetch the exact doc URL (overview + auth + reference) until specifics (URLs, scope …
- Distinctive: line 43
Use web search and official documentation tools available in the environment repeatedly until each memo subsection has support (or a stated …
- Distinctive: line 66
If the product does not use OAuth, state that and skip to A4.
- Distinctive: line 104
After the memo, add a short section for implementers: what layers exist, not how to code them.
“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Evaluate and scaffold a new Duct connector in two layers: (1) a Connector research memo backed by current official documentation, discovered through heavy use of web search (and fetching doc pages when snippets are thin); (2) a compact map of where wiring lives in this monorepo. Do not treat this skill as a line-by-line implementation guide—implementation starts only after the memo is reviewed.
When to use
- User names a product or API (e.g. Mixpanel, GA4 Data API, Google Search Console, Stripe Billing read-only).
- User wants go/no-go facts: auth model, scopes, quotas, prerequisites, and which endpoints support read-only reporting.
Usage
/add-connector "<VendorOrProduct>" [memo-slug]Examples:
/add-connector "Google Analytics 4 Data API"
/add-connector "Mixpanel" mixpanelIf the user does not ask for a file, deliver the Connector research memo in chat. If they ask to save it, write it as a dated record, docs/engineering/YYYY-MM-DD-<slug>-connector-research.md, with the author and date line every doc carries (docs/README.md, Naming).
Phase 0 — Web search protocol (mandatory)
Complete this before filling the memo. Prefer primary sources over blogs or forums.
- Start official: Search vendor developer docs (e.g. narrow to the vendor domain or
site:developers.google.comfor Google APIs). Use queries like"<product> API authentication","<product> OAuth scopes","<product> rate limits". - Follow up: Run additional searches for OAuth app verification / consent, quotas, pagination, API versioning, deprecation notices, and sandbox vs production.
- Cross-check freshness: Many analytics APIs version by year or have sunset dates—confirm the doc revision or announcement date; prefer the latest stable API name.
- Conflicts: If sources disagree, vendor documentation wins over Stack Overflow or generic tutorials.
- Deep pages: When search results are vague, open or fetch the exact doc URL (overview + auth + reference) until specifics (URLs, scope strings, limits) are confirmed.
Use web search and official documentation tools available in the environment repeatedly until each memo subsection has support (or a stated “not documented / N/A”).
Citations: Where possible, every subsection (A1–A7) should reference at least one official doc URL. Use full URLs in the memo.
Phase A — Connector research memo (main deliverable)
Produce markdown with these sections. Use today’s docs; note the API/product version if the vendor names one.
A1. Product fit for Duct
- What read-only metrics or entities would feed a brief/report (align with normalize → brief → synthesis).
- What is out of scope (write/mutation APIs, PII-heavy surfaces, enterprise-only gates).
A2. Authentication model (deep dive)
- Mechanism: OAuth 2.0 (authorization code + PKCE?), API keys, service accounts, signed JWTs, partner tokens, etc.
- User-vs-server: Suitable for browser OAuth (user connects, refresh token to app) like current Google Ads, or server secrets only?
- Token lifetime: Access vs refresh, rotation, revocation—note implications for client
sessionStoragevs future server-side token storage (describe current Duct pattern briefly: OAuth callback can redirect with tokens for the web app—seebackend/routes/auth.py—without mandating architecture changes).
A3. OAuth specifics (if applicable)
If the product does not use OAuth, state that and skip to A4.
- Grant types supported; which grant Duct should use for server-side reporting.
- Authorization and token endpoints (and JWKS if relevant).
- Scopes as full strings: minimum set for read-only reporting vs broader; incremental / additive consent if offered.
- Consent and verification: Internal vs public app, sensitive or restricted scopes, OAuth verification or security assessment requirements and timelines.
- Redirect URIs: HTTPS rules, localhost in dev, path matching—compatibility with a public API origin (e.g.
{API_PUBLIC_URL}/auth/connectors/{id}/oauth/callbackor vendor-required paths).
A4. Permissions, policies, and limitations
- Account model: Workspaces, properties, sites, MCCs—what the user must select after connect.
- Roles: Minimum role for API access; org policies that block APIs.
- Rate limits, quotas, pagination: Documented defaults; batch or async patterns.
- Data freshness and history: Latency, retention, backfill limits.
- Compliance / region: Data residency, PCI/HIPAA notes if relevant (e.g. payment data).
A5. Prerequisites and setup (operator checklist)
- Vendor-side: Projects, apps, API enablement, billing, domain verification, approvals.
- Duct-side (conceptual): Env var names you expect later in
backend/config.py(no implementation).
A6. APIs Duct can use (inventory)
From the official API reference, list:
- Reports/queries/endpoints suited to periodic briefs (dimensions, metrics, date ranges).
- Required identifiers (property id, site URL, account id, etc.).
- Official SDKs (e.g. Python) vs REST—maintenance tradeoff.
A7. Risks and open questions
- Doc gaps, deprecations, sunset dates.
- Sandbox/test behavior vs production.
Phase B — Duct repo structure (template only)
After the memo, add a short section for implementers: what layers exist, not how to code them. Google Ads is the reference implementation in-tree.
| Concern | Purpose | Typical locations |
| Connector registry | ConnectorMeta + adapter registration | backend/service/connectors.py; implementation under backend/service/...; ensure registration import in backend/server.py |
| OAuth entrypoints | Browser authorize/callback | backend/routes/auth.py (currently connector-specific branches for Google Ads) |
| Account listing API | GET /api/connectors/{id}/accounts when CAP_ACCOUNTS is set | backend/routes/connectors.py |
| Generate pipeline | Fetch, normalize, UnifiedReport.briefs | backend/routes/generate.py; pattern under backend/service/google/ |
| Typed brief contract | JSON shape consumed by the app | e.g. backend/service/google/schema.py for Ads |
| App surfaces | Connections UI, generate flow, report icons | app/src/app/(app)/connections/page.jsx, app/src/app/(app)/generate/page.jsx, app/src/lib/api.js, app/src/components/ReportsList.jsx |
| Site listing | Hub card on getduct.ai/integrations, and a page once its copy is written | scripts/integrations/connectors.py; scripts/build_integrations.py --check fails in CI until the connector is listed |
Template rule: The research memo drives whether you reuse a Google-family OAuth stack, implement list_accounts, add a new briefs.<connector_id> shape, and extend /api/generate. No code changes are implied until Phase A is approved.
Phase C — Handoff
Close with:
- Recommendation: feasible for Duct MVP, feasible with conditions, or not recommended—with reasons tied to memo sections.
- Suggested
connector_idslug (lowercase, snake_case, stable) if implementation proceeds. - Open questions requiring product or security review.
Notes
- Detailed backend conventions:
backend/AGENTS.md. - Google Ads remains the only fully wired connector; use it as the on-repo shape, not as a requirement that every connector use OAuth.
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
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.

