fix(data): remove duplicate node_id 243113 — one bad row disables the whole trusted-agents feed

[pstayets]

What

pilot-mom and pilot-director are two entries in trusted-agents.json for the same node — identical address (0:0000.0003.B5A9) and identical node_id (243113). This removes the redundant pilot-director entry (7-line deletion, nothing else). Node 243113 stays trusted through the pilot-mom entry; no node_id is invented.

How the collision got here

pilot-director was already in the list. Commit #31 ("Add pilot-mom, node 243113") added pilot-mom reusing that same node_id — introducing the duplicate. Since both entries point to the same node, #31 was effectively renaming the planner to pilot-mom (the name the docs, the app roster, the MCP package, and the homepage demo all use) but left the old pilot-director alias behind. This PR completes #31 by removing that stale alias.

Why it matters (not cosmetic)

Load() treats a duplicate node_id as fatal for the entire list — it returns before committing the parsed roster (data.go):

if other, exists := idx[a.NodeID]; exists {
    return fmt.Errorf("duplicate node_id %d ... %q and %q", ...)  // aborts the whole load
}
...
byNode = idx ; all = doc.Agents   // never reached

So this single collision makes all 438 entries fail to load on every daemon, on every hourly refresh:

WARN trustedagents fetch failed err="load: duplicate node_id 243113 in
     trusted-agents list: \"pilot-mom\" and \"pilot-director\""

(observed live on a fresh v1.13.2 daemon joining the backbone)

Consequences today:

• Field daemons (shipped v1.13.2, embedding the pre-Add pilot-mom (MOM) to the trusted-agents allowlist #31 snapshot that lacks pilot-mom) reject every refresh and stay frozen on that older embedded list — so pilot-mom / pilot-director are never trusted, and nothing added to the feed since is picked up.
• A binary built from current main embeds this colliding list, so init() hits the same error and sets byNode to an empty map — booting with zero trusted agents. TestEmbeddedListLoads doesn't catch this: it only asserts byNode != nil, and the empty map is non-nil.

Verification

• git diff = 7 deleted lines, only the pilot-director object + the preceding comma. No reformatting; trailing-newline state preserved.
• JSON valid; 437 agents; zero duplicate node_ids; pilot-mom present with node_id 243113; node 243113 still in the list.
• go test ./... → ok (full suite).

Scope

Minimal data change only. The two entries are byte-identical on address and node_id, so they are provably the same node and removing the alias is the correct fix (there is no distinct second node to assign a new id to). A companion PR (loader hardening) makes Load() degrade gracefully so a future duplicate can't drop the whole feed again.

🤖 Generated with Claude Code

https://claude.ai/code/session_0142ryqVGEmJN66VZtCC7wFD

[TeoSlayer]

Superseded — the duplicate pilot-director row was already removed on main via #34. Closing.