ops: activate six governed estate agents

[angellllkr-eng]

Adds six bounded agent contracts (Cartographer, Security Sentinel, Release Guardian, Reality Gate, Commercial Operator, A11-K Commander), a machine-readable manifest, and a scheduled GitHub Actions heartbeat/scope-check workflow. The automations are approval-first and cannot perform destructive, secret, DNS, billing, repository-delete or uncontrolled production mutations.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[angellllkr-eng]

P0 estate review: keep this PR approval-first and evidence-first. Before merge, reconcile the six-agent agent-manifest.yml with the broader AUTOMATION_MATRIX.md / ESTATE_AUTOMATION.yaml role set so there is one authoritative contract. Validate least-privilege permissions, schedule/concurrency behavior, secret fallback behavior, and the fact that the heartbeat is only an observation artifact. Do not weaken Reality Gate, commercial checks, or security checks to obtain green CI. Current mergeability is false; treat that as a blocker to resolve, not something to bypass.

[angellllkr-eng]

Status check (automated, read-only): this PR is currently not mergeable — GitHub reports mergeable_state: dirty, meaning there's a conflict against main on branch estate-reconstruction-2026-08-10. 14 commits, +448/-0 across 14 files, P0 labeled, review requested from @angelkrustevtopa-jpg but not yet submitted.

Next step: resolve the conflict on the branch (needs direct repo/code access — not doable via this metadata-only integration), then re-run mergeability check before merge.

[angellllkr-eng]

Review receipt — 2026-08-17 (agent)

Merge recommendation: DO NOT MERGE as-is.

State

• mergeable_state: dirty (conflicts with current main)
• Checks that ran: CodeQL + Analyze → success
• Branch: estate-reconstruction-2026-08-10 @ 7dd1d100
• 14 commits, +448 lines, 14 files

Why dirty / conflicts

Current main already contains a parallel (and evolved) agent layer:

• agents/registry.yml, agents/ESTATE_AGENTS.yaml, agents/*.agent.md (reality-gate, security-guard, commercial-journey, etc.)
• Multiple live workflows: estate-reality.yml, estate-security.yml, estate-discovery.yml, estate-cartographer.yml, org-automation.yml, etc.

This PR introduces a different naming set (agents/estate-cartographer.md, security-sentinel.md, …) + new workflows under root workflows/ and .github/workflows/agent-ops.yml. Direct merge would produce path and semantic conflicts.

Content quality

The contracts themselves remain sound and non-destructive:

• Explicit approval-first / no secret values / no blind delete / missing-evidence = UNKNOWN
• Six bounded agents + heartbeat + scope check align with the P0 in agent-control-plane#382

Recommended next step (owner)

1. Rebase / rewrite this PR against current main or close it and open a focused PR that only adds the missing pieces (manifest alignment + any contracts not already present on main).
2. Prefer a single source of truth for agent contracts (current main registry vs this PR’s six files).
3. After clean rebase, re-run the scope-check heartbeat and confirm it produces a receipt artifact.

No destructive action taken. Evidence only.

[angellllkr-eng]

Status sync with #382 (2026-08-17)

This PR remains not mergeable (dirty against current main).

Current main already carries an evolved agent layer (agents/registry.yml, multiple *.agent.md, live estate-* workflows). Merging the six-agent set from this branch would create path + semantic conflicts.

Recommendation (aligned with control-plane#382):

• Prefer close this PR and treat the main-branch contracts as the living source of truth, or
• Open a focused additive PR that only brings any missing contracts / heartbeat that are truly absent on main, without renaming or weakening gates.

Owner decision requested in the parent reconstruction issue. No action taken here beyond evidence.

[angellllkr-eng]

Follow-up on the mergeability check above.

Pulled the actual diff on ESTATE_AUTOMATION.yaml between this branch and main. This isn't a small textual conflict — main has a structurally redesigned version of the same config since this branch was cut:

	this branch	main
top-level key	subagents:	agents:
agent count/names	10 (cartographer, security, ci-guardian, reality-gate, release-guardian, integration-guardian, dependency-guardian, commercial-journey, recovery, archivist)	8 (discovery, security, reality, release, governance, commercial, evidence, owner_gate)
structure	flat scope/actions/schedule	adds class:, cron schedules, output: paths, routing:, states: lifecycle enum
unique to main	—	owner_gate (explicit approval gate for production/secret/billing/DNS/delete/org-permission changes), append-only evidence recorder, full state machine

main has also gained four files this branch doesn't have at all: SECRETS_INVENTORY.md, SUBAGENTS.md, CURRENT_ESTATE_MAP.md, ESTATE.yaml. This branch has one file (AUTOMATION_MATRIX.md) that isn't on main.

Read: this PR is superseded, not just conflicted. main's schema keeps the same safety principles (evidence-first, non-destructive, approval-required mutations) but is more mature — explicit owner authorization gate + state machine that this branch's version doesn't have. Merging this as-is would regress main to an older design and drop the four newer files.

Recommend: close this PR in favor of what's already on main, after confirming AUTOMATION_MATRIX.md (the one file unique to this branch) isn't needed — or cherry-pick just that file onto main separately if it is. Owner review requested before any action.