Make MCP Search-Result Ids Self-Contained Fetch Handles

A live ChatGPT retest of the 5-tool MCP surface (2026-06-09) showed the search→fetch journey failing on multi-source hosted packages. Search hits carried id = stream:recordid plus a SEPARATE connectionid field, so a model had to carry TWO values between tools. fetch(id) without connectionid returned a typed 409 ambiguousconnection; ChatGPT's rendered envelope buried the second field and its model never completed a fetch (retrying with both fields was verified to work). OpenAI's search/fetch contract treats result ids as single opaque handles; ours leaked a join requirement into the model loop.

Chosen grammar: {connectionid}/{stream}:{recordid} (e.g. cin4f2a/orders:o1), with the legacy stream:recordid form still parsed when no / is present.

Search result ids SHALL be single opaque handles sufficient to fetch the result with no additional argument. When a hit carries a source connection, the adapter SHALL encode it into the id using the grammar {connectionid}/{stream}:{recordid}. / is reserved as the connection separator because the adapter rejects / inside connection ids, stream names, and record ids, so a / unambiguously marks the self-contained form and every legacy stream:record_id id keeps parsing unchanged. The complete id SHALL appear in model-visible content[] text as well as structuredContent.results; an id SHALL NOT be truncated in text below usability as a fetch handle.