lumea-labs/polpo-skills · Archived

polpo-integrations

Build secure Polpo integrations using Connections, custom tools, trusted invocation context and hidden bindings, MCP servers, OAuth, and complete tool bundles.

First seen Aug 24, 2026

Installation

$ npx skills add lumea-labs/polpo-skills --skill polpo-integrations

Summary

  • Build secure Polpo integrations using Connections, custom tools, trusted invocation context and hidden bindings, MCP servers, OAuth, and complete tool bundles.
  • Use for credentials, tenant-scoped connection selection, tool packaging/execution, grants, scopes, or external service integration.

Stronger alternatives

This repository is archived — consider an actively maintained alternative.

Similar popular skills

Related neighbors and high-traction skills in the same topics — useful to compare before installing.

Also in this package

Other skills from lumea-labs/polpo-skills.

npx skills add lumea-labs/polpo-skills

Browse all from lumea-labs/polpo-skills

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Not declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Repository health

Default branch main
Open issues 0
Status Archived

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 2,285 B
  • docs SUMMARY.md 317 B

History

  1. First seen on skills.sh
  2. First recorded snapshot · 5 installs

SKILL.md

Polpo Integrations

Keep model-visible arguments separate from host-owned identity, authorization, and credentials. The model expresses intent; the runtime resolves trusted context and capabilities.

Workflow

  1. Decide whether the integration is a built-in tool, custom tool, MCP server, Channel handler,

or request-scoped client tool.

  1. Define the smallest model-visible JSON schema and separately define trusted bindings and

Connection slots.

  1. Resolve user, tenant, resource, scope version, and grants server-side before tool execution.
  2. Package the complete relative source graph and validate it before upload.
  3. Resolve only active Connections with all required scopes and an unambiguous trusted binding.
  4. Execute with bounded time, cancellation, sandbox policy, idempotency, and redacted trace data.
  5. Test denied scope, missing/ambiguous Connection, rotated credentials, timeout, retry, malformed

result, and package dependency failures.

Invariants

  • Connection IDs and credentials are not model-controlled tool arguments.
  • ToolInvocationContext is copied, JSON-validated, deeply frozen, and host-owned.
  • serverBindings can read only supported immutable invocation paths.
  • Hidden bindings and Connection capabilities never enter model history or serialized tool

arguments.

  • A tool cannot widen its granted scopes, filesystem access, or execution surface.
  • Packaging failure is detected before deployment; runtime 502 is not an acceptable dependency

discovery mechanism.

References

  • [references/custom-tools.md](references/custom-tools.md)
  • [references/trusted-context-and-bindings.md](references/trusted-context-and-bindings.md)
  • [references/connections.md](references/connections.md)
  • [references/mcp-and-oauth.md](references/mcp-and-oauth.md)
  • [references/packaging-and-runtime.md](references/packaging-and-runtime.md)
  • [references/contract-version.md](references/contract-version.md)