fix(telemetry): a second export is a second plugin

[blafourcade]

What

OpenCode loads every function-valued named export of a file in plugin/ as a plugin factory of its own. opencode-plugin.js exported journalCallFor beside AiddTelemetry as a spawn-free test seam; OpenCode called it with one argument, received null, and died reading .auth off it before any session started.

Installing this framework for OpenCode therefore made the tool it measures unusable.

Measured

Against opencode 1.14.20, in a freshly installed scratch project:

file in .opencode/plugin/	opencode run
export const Stub = async () => ({ event })	runs
+ export function helper(){ return null }	TypeError: null is not an object (evaluating 'S.auth')
+ export const helper = { a: 1 } (not a function)	runs
opencode-plugin.js as shipped	dies at load

The second half

The same run exposed the rest of the fault. OpenCode publishes session.created on its own bus and never delivers it to a plugin's event hook, and opencode run is always such a session. So the journal never received a session_start, never created the run file the rest of the session appends to, and dropped the turn-end and every task declaration after it — while telemetryLocalRead declared the tool covered and aidd telemetry read, which reads only sessions the run journal knows, could never find one.

The first call a session produces now opens it, carrying the directory that call was already going to use. Never a new guess: for a session no session.created named, that is the plugin's own init-time directory, exactly what journalCallFor already handed turn-end and tool-used.

What is left is only what session.created alone could have said, and the README now says it: on a server serving more than one directory, a session it never announced is journalled under the plugin's own init-time directory rather than its own. The guard that pinned the old limit sentence now pins both the new fact and what it still costs.

Verified

End to end on a scratch repository with an isolated sink — the real sink was never written.

A real opencode run produced the first OpenCode journal this project has ever recorded:

{"type":"session_start","at":"2026-09-06T05:29:27Z","tool":"opencode","vendor_id":"ses_…"}
{"type":"task_declared","at":"2026-09-06T05:29:27Z","path":"aidd_docs/tasks/…/opencode.md"}
{"type":"turn_end","at":"2026-09-06T05:29:29Z"}

and aidd telemetry read answered OpenCode: read (3 new of 3).

The session_start came from the task declaration, not from turn_end — the case the second new test covers.

Gates

• node --test "scripts/__tests__/**/*.test.js" — 380/380
• cli vitest — 3501/3501, run twice: after the code change, and again after the doc changes, since the golden build matrix embeds the shipped plugin file
• biome check on every changed file — clean
• check-markdown-links.js — 0 broken in 797 files
• CATALOG.md and docs/prompts-documentation.md regenerate unchanged

The six new tests were written first and were red for the reason they name before the change landed.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VWNxk63AGKkqE8HRqHLjGp