When discover_kaizen_projects returns {"projects": [...]} bound to
context.projects, for_each can iterate the list directly. Multi-key
summaries (e.g. repo SBOM bulk) remain unchanged.
Implement discover_kaizen_scheduled_repos and discover_kaizen_projects per
kaizen-agentic ADR-005 contract: State Hub roster, roster.yaml filter, schedule
validation, and prepare_command emission. Register kaizen/resolver/shell source
types with unit tests and runbook dry-run instructions.
Add safe action interpolation and for_each binding for rule fan-out, update the weekly SBOM definition, cover the new evaluation path, and reconcile activity-core scope/workplans for the State Hub sync.
The CUST-WP-0045 canary failed validation twice without leaving any
record of what the model actually returned. The warning logged only the
error message ($: missing required property 'summary'), not the JSON
shape that triggered it — so diagnosing required modifying code and
re-running. Log a 2KB preview of the offending raw output alongside the
error so the next failure of this shape is one grep away from diagnosis.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The CUST-WP-0045 daily triage canary on 2026-06-01 hit a BrokenPipeError
on the llm-connect side. Two 5-minute timeouts were racing:
- _ACTIVITY_TIMEOUT = timedelta(minutes=5) in workflows.py
- LLM_CONNECT_TIMEOUT_SECONDS default 300 in llm_client.py
The 10KB curated digest + max_depth:2 + JSON schema enforcement pushed
Claude past 5 minutes. Whichever timer fired first killed the httpx call;
the model's late response arrived to a closed socket.
Read _ACTIVITY_TIMEOUT from ACTIVITY_TIMEOUT_SECONDS env (default 900 —
15 minutes) so judgement-call activities have headroom for slow LLM runs.
Operators should also widen httpx via LLM_CONNECT_TIMEOUT_SECONDS=840 so
httpx still times out slightly before Temporal, preserving the
clean-error contract.
Tests: 120 passed, 1 skipped.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The state-hub resolver was calling GET /sbom/status?repo={slug}, which State
Hub does not expose. Real SBOM routes are /sbom/, /sbom/{slug},
/sbom/snapshots/, /sbom/snapshots/{id}, /sbom/ingest/, /sbom/report/licences/.
The weekly-sbom-staleness ActivityDefinition was passing params {repos: all}
and the resolver was reading params.get("repo_slug", ""), so the URL
collapsed to /sbom/status?repo= and 404'd. _fetch_json swallowed the error,
the rule context.repos.sbom_age_days > 30 evaluated against {} and never
matched, and the weekly SBOM check has been a silent no-op for as long as
the route mismatch has existed.
Resolver now supports two modes selected by params:
- single-repo: {repo_slug: foo} → GET /sbom/{foo}, returns
{repo_slug, last_sbom_at, sbom_age_days, has_sbom}
- bulk: {repos: all} → GET /repos/, computes per-repo age, returns the
worst repo's fields hoisted to the top of the result alongside
stale_count, total_count, worst_* fields, and the full per-repo list
Never-scanned repos get a 99999 sentinel age so threshold rules treat
them as very stale without forcing the rule to special-case None.
Hoisting the worst entry to the top preserves the existing rule
expression context.repos.sbom_age_days > 30 (and target_repo:
context.repos.repo_slug, though that field is a separate interpolation
gap tracked as ADHOC-2026-06-01-T02). The integration tests'
aspirational per-repo iteration model is left intact.
Live validation against State Hub on 2026-06-01:
- single: activity-core → 36 days since 2026-04-26 ingest
- bulk: 48 repos total, 46 stale (>30d), worst is info-tech-canon (never
scanned), rule expression evaluates True
Tests: 120 passed, 1 skipped.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Surfaced while bringing up the dev worker for the CUST-WP-0045 T06 cutover.
weekly-sbom-staleness fires its state-hub resolver with query
repo_sbom_status, which hits GET /sbom/status?repo=. State Hub does not
expose that route, so _fetch_json returns {} and the rule
context.repos.sbom_age_days > 30 silently no-ops. The weekly SBOM check has
been a no-op for as long as the route mismatch has existed. Logged as a
low-priority adhoc rather than promoting to a workplan because the resolver
and definition both need a one-line decision (single-repo vs fan-out), not
multi-phase design.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
