tortoise:file-finding

Ingest a research finding into the epistemic graph.

Which Tortoise are you on?

• Hosted (cloud): your agent reaches the graph through the Tortoise MCP server (tortoise_* tools) — no local database or SDK needed. If you installed the skills via the onboarding wizard, the MCP server is already wired. Configuring it by hand? Use the equivalent project-scoped .mcp.json entry: {"type": "http", "url": "https://api.premiselabs.co/mcp/", "headers": {"Authorization": "Bearer ${TORTOISE_API_KEY}"}} — that file is committable, so it carries the env reference, never the key. A project-scope server stays ⏸ Pending approval until you approve it once — start claude in the project and allow the prompt, or run /mcp.
• Self-hosted: you run your own FalkorDB + local MCP server (or the Python SDK). The MCP tool names below are the same; SDK callers use sdk.create_point(...), sdk.query(...), sdk.list_sources().

The steps below use the MCP tools, which work for both audiences when an MCP server is configured.

Steps

1. Accept finding content and source from the user or calling agent.
2. Call tortoise_create_point(kind='statement', content=<finding>, extractedFrom=<source_url>) to create the Point with provenance.
3. Call tortoise_list_sources() for discovery — identify what sources already exist in the graph.
4. Call tortoise_query(kind='statement') to find existing Points; post-filter by source to find Points from the same origin.
5. If existing Points found, surface them as potential connection candidates.
6. Offer to create IMPL/NAND connections or flag for review.

Quality Gates

• G1 (Static): Content must be non-empty. Source must be a valid identifier (URL or name).
• G2 (Semantic): If the finding is near-identical to an existing Point, warn about potential duplicate.

Error Handling

• If tortoise_create_point fails, report the error and do not proceed to query.
• If tortoise_query returns empty, report "no existing Points from this source" — this is a success state.