test: use event-driven AFK injection waits

[JTInventory]

Intent

Reduce Firstmate behavior-test runtime without weakening assertions or gates by replacing fixed afk injection sleeps with bounded event-driven waits

What Changed

• Captain, replaced fixed AFK injection sleeps with bounded polling helpers that wait for the specific daemon/log state and print focused diagnostics on timeout.
• Added stable event waits for swallowed-Enter and digest delivery paths so the e2e verifies one clean digest and an empty escalation buffer after delivery.
• Updated CONTRIBUTING.md to describe the AFK injection e2e as event-driven single-digest coverage.

Risk Assessment

✅ Low: Captain, the change is limited to one AFK e2e test file and the event-driven waits preserve the original assertions while adding a stability window for duplicate detection.

Testing

The configured full-suite baseline was already green; I then reran the focused AFK injection e2e with tmux 3.4, verified the PR target guard, captured reviewer-visible CLI evidence for all three AFK injection scenarios, and recorded that the target event-driven run finished in 19.88s versus the base test's 26s fixed-sleep floor.

Evidence: AFK injection e2e transcript

cwd: /root/.no-mistakes/worktrees/76db9585a6d5/01KWF2MXFAGQAESNS0PV539AGE
head: bc82f3499ad7ccdd700445672b244da5482925f7
branch: ## HEAD (no branch)
tmux: tmux 3.4
command: bash tests/fm-afk-inject-e2e.test.sh
ok - Scenario A: partial input defers injection; digest arrives clean after idle
ok - Scenario B: swallowed Enter produces exactly one clean digest
ok - Scenario C: a normal captain status injects exactly one clean single-line sentinel digest
all e2e injection tests passed
real 19.88
user 0.77
sys 1.51

Evidence: AFK runtime evidence

intent: Reduce Firstmate behavior-test runtime without weakening AFK injection assertions or gates by replacing fixed sleeps with bounded event-driven waits.

target_e2e_runtime:
command: bash tests/fm-afk-inject-e2e.test.sh
ok - Scenario A: partial input defers injection; digest arrives clean after idle
ok - Scenario B: swallowed Enter produces exactly one clean digest
ok - Scenario C: a normal captain status injects exactly one clean single-line sentinel digest
all e2e injection tests passed
real 19.88

base_fixed_sleep_floor:
254:  sleep 6
272:  sleep 6
324:  sleep 8
371:  sleep 6
sum: 6 + 6 + 8 + 6 = 26 seconds before setup/teardown

target_event_waits:
345:  wait_for_event "Scenario A pending-input defer" 60 scenario_a_deferred_with_pending_input
363:  wait_for_event "Scenario A digest after idle" 60 scenario_a_delivered_after_idle
415:  wait_for_stable_event "Scenario B one digest after swallowed Enter" 90 12 scenario_b_delivered_once_after_swallowed_enter
462:  wait_for_stable_event "Scenario C one digest" 70 12 scenario_c_delivered_once

Pipeline

Updates from git push no-mistakes

✅ **intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

🔧 **Review** - 1 issue found → auto-fixed (2) ✅

• ⚠️ tests/fm-afk-inject-e2e.test.sh:259 - Captain, this wait now returns as soon as one digest is logged and the buffer is empty. The exact-count assertion runs immediately after, so a duplicate emitted on the next watcher or housekeeping cycle can arrive after the test already passed. Keep the event-driven wait, but add a short stability check that requires exactly one digest for at least one poll cycle after first delivery.

🔧 Fix: Stabilize AFK duplicate waits
1 warning still open:

• ⚠️ tests/fm-afk-inject-e2e.test.sh:408 - Captain, the stability window is counted inside the overall timeout, so Scenario B only allows the digest to first appear by about 7.8s even though the old fixed wait allowed about 8s before asserting. Increase the timeout by at least stable_ticks or change wait_for_stable_event so the stability clock starts after first success; otherwise this can add CI flake while trying to make the test more event-driven.

🔧 Fix: Captain, stabilize AFK wait clocks
✅ Re-checked - no issues remain.

✅ **Test** - passed

✅ No issues found.

• bash bin/fm-no-mistakes-pr-target-guard.sh || exit 1; 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"
• Provided baseline already passed: bash bin/fm-no-mistakes-pr-target-guard.sh || exit 1; 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".
• /usr/bin/time -p bash tests/fm-afk-inject-e2e.test.sh with transcript captured to /tmp/no-mistakes-evidence/01KWF2MXFAGQAESNS0PV539AGE/fm-afk-inject-e2e-transcript.txt.
• command -v tmux && tmux -V && bash bin/fm-no-mistakes-pr-target-guard.sh.
• Compared target event-driven waits against base fixed sleeps with git show 528ca21e3fc540cfbff6849eff8385bde8f878ee:tests/fm-afk-inject-e2e.test.sh | grep -n "sleep 6\|sleep 8" and grep -n "wait_for_event \|wait_for_stable_event " tests/fm-afk-inject-e2e.test.sh, recorded in /tmp/no-mistakes-evidence/01KWF2MXFAGQAESNS0PV539AGE/afk-runtime-evidence.txt.
• git status --short --ignored=no confirmed no working-tree artifacts were left behind.

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.