Repository navigation
Description
Problem
ADR-0017's launch table has three named rows and an escape hatch. Only two of
those rows have ever completed a turn:
| Name | Standing today |
|---|---|
claude |
Verified 2026-09-07, fresh and resumed |
hermes |
Verified 2026-09-07 fresh; resumed is #448 |
opencode |
Unverified — not installed on the machine #433 or #434 was written on |
| anything else | Works, as typed, with no row and no name |
The ADR also names three protocol-compatible Harnesses that are reachable
today through the escape hatch and have simply never been smoked:
- Grok Build —
grok agent stdio - GitHub Copilot CLI —
copilot --acp --stdio(public preview) - Gemini CLI —
gemini --acp
"Reachable through the custom value" is not the same as supported. A user who
types one of those gets no row, no documentation, and no evidence anyone has
run it. That is the gap this closes.
The bar is already written
ADR-0017 sets it: "A row moves in this table when that command exits zero."
scripts/probe-harness.sh attaches the configured Harness with no overlay,
runs one fixed prompt, and prints the handshake, the stop reason, the reply,
and whether the reply parsed as a Behavior proposal.
So this is not a design task. It is: install each Harness, run the probe,
record what initialize advertised, and move the row.
Scope
For each of opencode, Grok Build, GitHub Copilot CLI and Gemini CLI:
- Run
scripts/probe-harness.shand get exit zero on a fresh session. - Run it again on a resumed session, because that is where
hermesbroke
(fix(harness): A session/load that succeeds without restoring pins every turn to a dead session #448) and where the fake agent's error-shaped refusal hid the bug. - Record what its
initializeadvertises —loadSession,
mcpCapabilities.http,authMethods,sessionCapabilities,
promptCapabilities— because ADR-0017 already shows these differ enough
to matter, and Attaching real harnesses: per-harness MCP client config for Claude Code, opencode, Hermes, Pi, and Grok #166 depends on the MCP half. - Give it a named row in the launch table with its command and standing.
A Harness that fails gets a row too, saying so. An honest "does not work yet,
here is what it did" is worth more than absence.
Constraints from ADR-0017, not to be relitigated here
- No auto-approve flags, ever. None of the always-approve, yolo or
setSessionModeflags is passed by default. The Chat surface owns
permissions. - Gemini needs care. It has open issues about stdout pollution corrupting
its NDJSON. Smoke it before believing it. - Pi stays deferred. It has no first-party ACP and the research says it
wants a second adapter, not a compromise. Reopening that is a separate
argument. - A vendor extension we do not know (
x.ai/*and the like) gets
method-not-found, which the protocol allows.
Also unproven, and worth folding in
Step 5 of #434 never ran: every probe found no ai-buddy-mcp binary beside the
app and had AI_BUDDY_MCP_BIN unset, so no session ever got an MCP server and
no Harness ever called the speak tool. The stdio server is built and tested
on its own; that it is reachable through a session is still untested for
every Harness including the two verified ones.
Acceptance
- Each Harness above has a named row with its command and a dated standing
- Each row's standing was earned by a probe run, not by reading docs
- Fresh and resumed sessions both attempted, and the difference recorded
- What each advertises in
initializeis written down where Attaching real harnesses: per-harness MCP client config for Claude Code, opencode, Hermes, Pi, and Grok #166 can use it - The MCP path is exercised through at least one real Harness session
Related
#166 (per-harness MCP client config), #448 (resumed-session bug), #434
(closed — the first verification pass), ADR-0017.
🤖 Generated with Claude Code
Metadata
Metadata
Assignees
Labels
Projects
- StatusShow more project fieldsDone
Activity
Research: OpenAI Codex CLI as ACP Harness
feat(harness): Add Codex named Harness row (ACP)
omesser commentedon Sep 8, 2026
Codex named row shipped in #556 (merge
087b55e). Per that PR / #554, the smoked-turn gate for Codex is still outstanding (AI_BUDDY_HARNESS=codex scripts/probe-harness.sh). Leaving this epic open.— Cursor agent (Coder), on @omesser's behalf.
omesser commentedon Sep 9, 2026
Taking the next named-row slice: smoke
opencode(installed here; still unmarked in the launch table). Copilot CLI and Gemini CLI are not on this machine; Grok Build can follow in a later turn.— Cursor agent (Coder), on @omesser's behalf.
omesser commentedon Sep 9, 2026
opencodesmoke (2026-09-09)AI_BUDDY_HARNESS=opencode scripts/probe-harness.shtwice. First attach was fresh (the probe dir still named aclaudesession, sosaved_iddid not load). Second attach reusedses_f783d375bffe961nB046ND5aG5. Both exited 0 (end_turn).Handshake (both runs)
OpenCodeloadSession: trueai-buddy --mcp-stdio(the probe does not bind the loopback listener)authMethods:Login with opencode(advertised while signed in; the turn still ran)ses_f783d375bffe961nB046ND5aG5Turn (fresh and resumed)
Reply with exactly this one line and nothing else: Wave | Hello from the probe.end_turnWave | Hello from the probe.speakwas not called; the probe line came back as the turn text. MCP was the stdio shimcargo runalready hands, not a separately builtai-buddy-mcp. No code change: attach, resume, and parse all worked.Standing is in README and ADR-0017's opencode cell. Grok Build, Copilot CLI, and Gemini CLI are still unverified.
— Cursor agent (Coder), on @omesser's behalf.
omesser commentedon Sep 9, 2026
What
initializeadvertises, for acceptance bullets 3 and 4Two of the five fields this issue names —
sessionCapabilitiesandpromptCapabilities— have no column in the README table and do not belong in an ADR (docs/agents/docs.mdkeeps verification detail on the Issue or PR). Recording them here so #166 can read them without digging through a merged PR.Read by hand off one
initializerequest per Harness, with nosession/newand no flags.loadSession,mcpCapabilitiesandauthMethodsare also in the README capability table; the last two columns are only here.loadSessionmcpCapabilitiessessionCapabilitiespromptCapabilitiesagentInfoopencode1.18.30http,ssehttp,sse_meta.agentVersion1.0.25Grok Build also sends vendor extensions under
_meta(x.ai/hooks,x.ai/fs_notify,x.ai/capabilities) and anauthcapability object. Nothing in ai-buddy reads any of them, and an unknown vendor method gets method-not-found, which the protocol allows.Both were smoked with
scripts/probe-harness.sh, fresh and resumed, exit zero. #587 has the transcripts and moves the rows; #585 recordedopencode's standing.Still open here: GitHub Copilot CLI and Gemini CLI, neither installed on this machine, and the MCP bullet — no Harness has yet called
speakthrough a session, because a probe binds no loopback listener and so hands every Harness the stdio shim.— Claude Code agent (Coder), on @omesser's behalf.
🤖 Generated with Claude Code