feat(omp): add Oh My Pi harness adapter

.agents/skills/afk/SKILL.md

@@ -80,7 +80,7 @@ explicit word - the daemon just batches the notification.
 
 The daemon prefixes every injection with `FM_INJECT_MARK` (U+2063 INVISIBLE SEPARATOR), which has no normal keyboard keystroke and survives terminal transport as UTF-8 text.
 This is how firstmate tells a daemon escalation apart from a real message in the same pane.
-The marker travels with the message text; it does not rely on harness-level typed-vs-injected detection, which is not portable across claude, codex, opencode, pi, and grok.
+The marker travels with the message text; it does not rely on harness-level typed-vs-injected detection, which is not portable across claude, codex, opencode, pi, grok, and omp.
 
 ## Busy-guard and composer guard
 

.agents/skills/firstmate-orca/SKILL.md

@@ -13,7 +13,7 @@ It does not replace `AGENTS.md`, `docs/orca-backend.md`, or `harness-adapters`.
 
 Orca is a runtime backend, not an agent harness.
 The runtime backend owns the task endpoint and, for Orca, the task worktree.
-The harness is the agent process launched inside that endpoint, such as `claude`, `codex`, `opencode`, `pi`, or `grok`.
+The harness is the agent process launched inside that endpoint, such as `claude`, `codex`, `opencode`, `pi`, `grok`, or `omp`.
 Load `harness-adapters` for harness-specific launch, interrupt, resume, trust-dialog, and skill-invocation facts.
 
 Implementation details, metadata fields, teardown guarantees, limitations, and smoke evidence live in `docs/orca-backend.md`.

.agents/skills/harness-adapters/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: harness-adapters
-description: Agent-only reference for firstmate harness operations. Use before spawning or recovering a crewmate or secondmate, handling a trust dialog, sending a harness-specific skill invocation, interrupting or exiting an agent, resuming an exited agent, or verifying a new harness adapter. Contains verified facts for claude, codex, opencode, pi, and grok.
+description: Agent-only reference for firstmate harness operations. Use before spawning or recovering a crewmate or secondmate, handling a trust dialog, sending a harness-specific skill invocation, interrupting or exiting an agent, resuming an exited agent, or verifying a new harness adapter. Contains verified facts for claude, codex, opencode, pi, grok, and omp.
 user-invocable: false
 metadata:
   internal: true
@@ -44,6 +44,7 @@ If the captain asks for a new harness, propose verifying it first: spawn a trivi
 On `unknown`, ask the captain instead of guessing.
 A captain override always beats detection.
 When verifying a new adapter, record its env marker and command name in `bin/fm-harness.sh`.
+The `omp` (Oh My Pi) adapter's env marker is `OMPCODE=1`, set by omp for its child/tool processes; `bin/fm-harness.sh` checks it BEFORE the claude `CLAUDECODE=1` marker because omp sets BOTH, so omp is never misdetected as claude. Ancestry fallback also matches an `omp` command name or a `bun`/`node`/`python` process whose args reference omp.
 
 For stuck recovery, the target window's harness is recorded as `harness=` in `state/<id>.meta`.
 Use that value for interrupt, exit, resume, and skill-invocation facts.
@@ -52,15 +53,15 @@ Use that value for interrupt, exit, resume, and skill-invocation facts.
 
 Every verified primary harness has an empirically validated hook path for the "no turn ends blind" guard.
 `claude` and `codex` block directly through Stop hooks that preserve exit status 2 and stderr from `bin/fm-turnend-guard.sh`.
-`opencode`, `pi`, and `grok` expose passive lifecycle callbacks for this purpose, so their tracked primary adapters force one bounded follow-up or resume when the shared predicate blocks.
+`opencode`, `pi`, `grok`, and `omp` expose passive lifecycle callbacks for this purpose, so their tracked primary adapters force one bounded follow-up or resume when the shared predicate blocks.
 The exact hook files, commands, validation transcripts, scoping rules, and fail-open tradeoffs are owned by `docs/turnend-guard.md`.
 When changing any primary turn-end hook, validate the real harness behavior in a scratch project or throwaway home before trusting it, then update that doc and the relevant concise fact below.
 
 ## Primary pre-arm (PreToolUse) seatbelt
 
 Every verified primary harness also has a wired PreToolUse-equivalent hook that denies a watcher-arm anti-pattern (shell `&`, truncating pipe, bundling, broad `pkill -f fm-watch`) before it runs.
 `claude` and `codex` block directly through PreToolUse hooks; `grok` blocks the same way but requires every `$VAR` reference in its hook `command` string to carry an inline `:-default` or it fails to launch the hook entirely.
-`opencode` and `pi` block by throwing from `tool.execute.before` / returning `{block: true}` from `tool_call`.
+`opencode`, `pi`, and `omp` block by throwing from `tool.execute.before` / returning `{block: true}` from `tool_call`.
 The exact hook files, commands, output-shaping quirks (Claude Code only honors the deny when stdout is empty), and validation transcripts are owned by `docs/arm-pretool-check.md`.
 When changing any primary PreToolUse hook, validate the real harness behavior in a scratch project before trusting it, then update that doc.
 
@@ -72,6 +73,7 @@ Claude and Grok use tracked background-notify cycles around `bin/fm-watch-arm.sh
 Codex uses bounded foreground checkpoints through `bin/fm-watch-checkpoint.sh` because Codex cannot reason while a foreground tool call is running.
 OpenCode uses `.opencode/plugins/fm-primary-watch-arm.js`, which coordinates with the turn-end guard plugin and wakes the TUI with `client.session.promptAsync`.
 Pi uses the tracked `.pi/extensions/fm-primary-turnend-guard.ts` plus the tracked `.pi/extensions/fm-primary-pi-watch.ts`, both project-local extensions Pi auto-discovers once trusted.
+omp (a Pi fork) uses the tracked `.omp/extensions/fm-primary-turnend-guard.ts` plus the tracked `.omp/extensions/fm-primary-omp-watch.ts`, both project-local extensions omp auto-discovers once trusted (omp scans `.omp/extensions/`, never `.pi/`).
 When changing any primary watcher adapter, update `docs/supervision-protocols/`, `docs/turnend-guard.md` if a shared idle or turn-end hook changed, and the relevant concise fact below.
 
 ## Launch profile axes
@@ -96,6 +98,7 @@ The supported launch-profile flags below are verified locally; each row records
 | grok | `--model <model>` | `--reasoning-effort <low\|medium\|high>` | Verified on grok 0.2.99 (2026-07-13). `--effort` is an alias, but firstmate's profile axis is reasoning effort. As of 0.2.99 the ceiling is `high`; both `xhigh` and `max` are rejected with `use one of: high, medium, low`, so firstmate omits them. |
 | pi | `--model <model>` | `--thinking <low\|medium\|high\|xhigh\|max>` | Verified 2026-07-13 on Pi 0.80.6. `pi --help` advertises `off`, `minimal`, `low`, `medium`, `high`, `xhigh`, and `max`; `pi --print --model openai-codex/gpt-5.6-sol --thinking max 'Reply with exactly OK.'` completed successfully. |
 | opencode | `--model <provider/model>` | none for firstmate's interactive launch | Verified on opencode 1.17.6. `opencode run` has `--variant`, but firstmate launches the interactive `opencode --prompt` path, which has no verified effort flag. |
+| omp | `--model <model>` | `--thinking <low\|medium\|high\|xhigh>` | Verified live 2026-07-10 (omp v16.3.15). omp is a Pi fork: `--thinking` also accepts `off\|minimal\|auto` but not `max` (omit it), `--auto-approve` grants autonomy, `-e/--extension` loads the turn-end/watch supervisors. |
 
 When a requested effort value is outside the harness-specific accepted set, `fm-spawn` records the requested `effort=` in meta but emits no effort flag for that harness.
 This preserves launch success instead of passing a known-bad value.
@@ -109,6 +112,7 @@ Natural language is acceptable if uncertain.
 - codex: `$<skill>`, for example `$no-mistakes`; `/<skill>` is claude-only and codex rejects it as "Unrecognized command".
 - opencode: no separate verified skill invocation beyond normal slash-command behavior; use natural language if the exact skill command is uncertain.
 - pi: no separate verified skill invocation beyond normal command behavior; use natural language if the exact skill command is uncertain.
+- omp: `/no-mistakes` (omp supports slash-command skills, like claude); no separate verified invocation quirk, use natural language if uncertain. The skill must be discoverable by omp (a scope omp reads, e.g. `~/.omp/agent/skills/` or the project `.claude/skills/`).
 - grok: `/<skill>`, for example `/no-mistakes` (same form as claude). Verified end to end: grok discovers the user-level `no-mistakes` skill, `/no-mistakes` invokes it, and grok drives a real `no-mistakes axi run`. Like codex's `$`/`/` popups, typing `/<skill>` opens grok's slash-autocomplete, so a too-fast Enter selects the popup entry instead of sending, and for an argument-taking command (like `/no-mistakes`'s optional task-first argument) that first Enter only expands the popup selection into an argument-hint placeholder rather than submitting - a genuine second Enter is required (see the grok section below for the 2026-07-03 incident and fix). `fm_tmux_submit_core`'s retried Enter (used by `fm-send` on the tmux backend) already handles this correctly by reading the cursor row; the herdr backend needed a dedicated fix (`fm_backend_herdr_composer_state`, docs/herdr-backend.md) because its prior delta-based verification false-positived on that same popup-close content change.
 
 ## claude (VERIFIED)
@@ -275,3 +279,30 @@ The adapter therefore runs the shared predicate and, when it returns 2, forces o
 It does not pass `--permission-mode`, so the passive hook cannot escalate the primary session's tool permissions.
 Project-local Grok hooks require folder trust, verified with launch-time `--trust`; if the primary firstmate checkout is not trusted for Grok hooks, this primary guard fails open and `fm-guard.sh` remains the next-command alarm.
 Grok's primary watcher protocol is Claude-shaped background-notify around `bin/fm-watch-arm.sh`; the passive Stop hook is only a backstop for blind turn ends.
+
+## omp (VERIFIED 2026-07-10, omp v16.3.15)
+
+omp (Oh My Pi, https://omp.sh) is a Pi fork with a Pi-compatible extension API (`turn_end` / `tool_call` events, `{block:true}` from `tool_call`, `pi.sendUserMessage(..., { deliverAs: "followUp" })`). Ported from the pi adapter and live-verified 2026-07-10 (see the validation record below): detection, extension load, the turn-end guard + watcher re-arm loop, and crew dispatch on omp are all confirmed on a real session. A few narrow items remain unexercised and are flagged inline (seatbelt deny, tmux-backend busy signature, exit/interrupt keys).
+
+| Fact | Value |
+|---|---|
+| Env marker | `OMPCODE=1` (omp also sets `CLAUDECODE=1`; `fm-harness.sh` checks `OMPCODE` first) |
+| Busy-pane signature | herdr backend: native busy-state verified working 2026-07-10. tmux backend: `Working...`/`Working…` / `esc to interrupt` regex still PENDING a tmux-backend run; override with `FM_BUSY_REGEX` / `FM_COMPOSER_IDLE_RE` |
+| Exit command | `/quit` - VERIFIED 2026-07-10 by `tests/fm-omp-primary-live-e2e.test.sh` (clean `OMP_EXIT=0`; watcher + arm children reaped on exit) |
+| Interrupt | single Escape (inherited from pi; not exercised by the e2e) |
+| Skill invocation | `/<skill>` (e.g. `/no-mistakes`) |
+| Autonomy | `--auto-approve` (omp HAS an approval system, unlike pi; fm-spawn passes it for crewmates - verified live 2026-07-10: two crewmates ran unattended and shipped PRs) |
+
+omp is claude-compatible (sets `CLAUDECODE=1`) but does NOT implement Claude Code's `.claude/settings.json` `Stop`/`PreToolUse` event-hook contract - it uses its own `.omp/extensions/` runtime instead. That is exactly why detection must resolve `omp`, not `claude`: a claude-detected omp session would install `.claude/settings.json` Stop/PreToolUse hooks that omp never fires, silently disabling supervision.
+
+**Primary-session guard (VERIFIED live 2026-07-10).**
+The primary's turn-end guard AND PreToolUse seatbelt both live in `.omp/extensions/fm-primary-turnend-guard.ts`. It listens for `turn_end` because OMP has no `agent_settled` event (Pi 0.80.5-only). The `guardFollowupActive` one-shot skip suppresses the guard on its OWN injected follow-up turn, but - unlike pi's per-logical-run `agent_settled` - it re-nags on each subsequent blind turn until supervision is armed (the live e2e saw 3 injections before the model armed the watcher). On block it `await`s `pi.sendUserMessage(..., { deliverAs: "followUp" })` when `bin/fm-turnend-guard.sh` returns 2. The seatbelt returns `{ block: true }` from the `tool_call` handler when `bin/fm-arm-pretool-check.sh` denies a bash command.
+
+**Primary watcher (VERIFIED live 2026-07-10).**
+`.omp/extensions/fm-primary-omp-watch.ts` registers the `fm_watch_arm_omp` tool (primary path, called instead of a foreground bash arm) plus the `/fm-watch-arm-omp` command as a human fallback (the command notifies via `ctx.ui.notify`). Arming spawns `bin/fm-watch-arm.sh --restart` attached to the live omp process and sends a follow-up wake when the child exits with an actionable reason; a one-shot `process.once("exit")` listener (mirroring Pi #397) plus `session_shutdown` stop the arm child on exit. `bin/fm-session-start.sh` reports when the running omp session has not loaded both extensions (markers `state/.omp-turnend-extension-loaded` and `state/.omp-watch-extension-loaded`). Both are project-local `.omp/extensions/*.ts` files omp auto-discovers once the project is trusted (approve trust once per clone, or launch with `-e` as the trust-free fallback). The tool schema uses `pi.zod.object({})` (OMP-canonical) rather than Pi's typebox `Type.Object({})`, and OMP's ToolDefinition has no `promptSnippet`/`promptGuidelines` fields.
+
+**Live validation record, 2026-07-10 (omp v16.3.15, herdr backend).**
+A fresh `omp` in the firstmate home became the first mate (root `AGENTS.md` loaded via the `agents-md` provider), `bin/fm-harness.sh` and `bin/fm-lock.sh` both detected `omp`, and both `.omp/extensions/` loaded (markers written). It dispatched two crewmates (`fm-webull-broker-w7`, `fm-finnhub-free-f3`) - each a real omp session in its own treehouse worktree, running autonomously under `--auto-approve` - which shipped two green PRs. The turn-end guard fired on a real multi-turn primary ("TURN WOULD END BLIND ... 2 task(s) in flight, but no live watcher holds this home lock") and the primary re-armed the watcher (alive) instead of ending blind.
+
+**Automated live E2E, 2026-07-10 (omp v16.4.0, tmux).**
+`tests/fm-omp-primary-live-e2e.test.sh` (opt-in, `FM_OMP_LIVE_E2E=1`; mirrors `fm-pi-primary-live-e2e.test.sh`) passes: it launches omp on a private tmux socket with the tracked extensions via `-e`, drives bash/read turns, triggers the turn-end guard, arms `fm_watch_arm_omp`, delivers a watcher wake, drains + re-arms, and asserts one-or-more guard injections (omp re-nags per blind turn), NO foreground `bin/fm-watch-arm.sh` arm, a live re-armed watcher pid, and a clean `/quit` (`OMP_EXIT=0`) that reaps both the watcher and arm children. It uses the default (already-authed) agent dir because a fresh `PI_CODING_AGENT_DIR` triggers omp's blocking first-run setup wizard. Still not exercised anywhere: a seatbelt `{block:true}` deny of an arm anti-pattern, the tmux busy-footer regex (`FM_BUSY_REGEX`), and the interrupt (Escape) key.

.no-mistakes.yaml

@@ -23,9 +23,12 @@ disable_project_settings: true
 # iterate every tests/*.test.sh, run each, and fail the step if any one exits
 # non-zero (an agent-driven test step has crashed the daemon). The e2e tests need
 # tmux on PATH, which the firstmate environment provides.
+# Each test runs with ambient harness/backend markers stripped
+# (OMPCODE/CLAUDECODE/HERDR_ENV/TMUX) so the gate matches CI's clean env even when
+# the daemon or push originates from inside an omp or herdr session.
 commands:
   lint: 'bin/fm-lint.sh'
-  test: 'command -v tmux >/dev/null || { echo "tmux is required for e2e tests" >&2; exit 1; }; tmux -V; rc=0; for t in tests/*.test.sh; do echo "== $t =="; bash "$t" || rc=1; done; exit "$rc"'
+  test: 'command -v tmux >/dev/null || { echo "tmux is required for e2e tests" >&2; exit 1; }; tmux -V; rc=0; for t in tests/*.test.sh; do echo "== $t =="; env -u OMPCODE -u CLAUDECODE -u HERDR_ENV -u TMUX bash "$t" || rc=1; done; exit "$rc"'
 
 # Keep test evidence out of this repo; it stays in a temp dir instead.
 test: