EnvironmentFile + direct venv ExecStart skipped k8s:// resolution and
could not parse REPO_MAP JSON. Wrapper loads env safely and resolves
ClusterIPs; PATH includes kubectl.
Harden the daily FI prompt and enforce minimum signal so green runs are
not empty templates. Raise default max_tokens. Push origin after commit
when enabled so workstation/Forgejo see briefs (diverged repos still OK).
activity-core owns schedule and spawn; rein-aharness executes the research
brief via llm-connect, commits the artifact, and posts fi_daily_brief for
fi_brief_status idempotence — same layering as brief-daily for Binky.
Maintainer decision, 2026-07-29: adopts TRSL V1C1 as this repo's
preliminary governing license, per target-revenue's
workplans/TREV-WP-0008-governance-and-pilot-rollout.md T05. Full
specialist legal review is deferred until out of beta (target-revenue
SCOPE.md §1). No Phase is yet declared for this repo.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Full cutover executed on railiance01 with operator go-ahead: host
secrets dir + checkout renamed, host venv recreated from scratch
(renaming a venv directory breaks its embedded shebang paths), image
rebuilt/imported, renamed k8s manifests applied alongside the old
namespace (not overwriting it), rollout + in-cluster smoke verified,
then the authoritative host smoke script run against the live
deployment: ok: true, committed: true, pushed: true, with a matching
harness_smoke event confirmed in State Hub. Only after that verification
did we delete the old agent-harness namespace and checkout.
Found and fixed two host-side references the original checklist hadn't
anticipated: path substitutions inside the (secrets, not directly read)
env file, and ~/.ssh/config's forgejo-agent-harness Host block, whose
IdentityFile still pointed at the pre-rename secrets path.
HARNESS-WP-0002 is now fully done (4/4).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude Code executes its own tools internally in --print mode -- there
is no way for a caller to externally dispatch individual tool calls
without abandoning that self-contained agent model. What
--output-format stream-json --include-hook-events does allow: observing
each tool_use/tool_result/hook event in real time.
- adapter.py: AgenticClaudeCodeAdapter gains an optional on_tool_event
callback; streaming mode (Popen + background reader thread) is used
only when set, blocking subprocess.run path is unchanged otherwise.
- runner.py: run_task gains emit_tool_events/on_tool_event, collecting
events onto RunResult.tool_events and posting a tool_call hub event
per tool when reporting is enabled.
- cli.py: --stream-tool-events flag on `run`, prints each event as a
tagged JSON line ahead of the unchanged final result block.
Live-verified against the real claude CLI: 5 real tool events streamed
correctly (2x Bash, 1x Write) plus Stop hook lifecycle events, real
commit landed, final result block unchanged. 13 new tests
(test_adapter.py + 2 in test_runner.py), all passing.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
agent_harness -> rein_aharness (package + all imports), CLI command
agent-harness -> rein-aharness, Docker image tag, k8s namespace/labels/
names, Makefile targets, deploy script env var/paths. In-repo identity
strings (hub event source, metrics harness field, default assignee,
argparse prog name, commit author identity) updated to match.
Historical documents left untouched on purpose: docs/adr/ADR-001-agent-harness-architecture.md,
docs/architecture.md (dated v0.1 snapshot), workplans/HARNESS-WP-0001
(completed under the old name), and the SSH host alias
"forgejo-agent-harness" (external ~/.ssh/config entry, not owned here).
Verified: 47/47 tests pass, CLI runs correctly from a fresh venv,
`make image` builds and the resulting container runs correctly.
deploy/README.md gained an explicit rename cutover checklist for what
this session cannot safely do unattended -- moving the host-side
secrets dir and checkout on railiance01, and not deleting the old k8s
namespace until the new one is confirmed working. The actual live
cutover (running that checklist against the real Railiance deployment)
is not attempted here -- real production surgery on binky-control's
live automation, needs the operator present.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The whole-run-as-one-dispatch_tool-call wrapper is live-proven now; the
per-tool-call hook refactor this task describes remains open.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
