Adopt documentation standard v1.0.0

Adopt documentation standard v1.0.0 Takes the reviewed release candidate, with four corrections and the release gate cleared locally. What the candidate adds. Repository declarations gain visibility, so client work can be classified rather than assumed, and on-call repositories declare an alerts register that the audit checks against the runbook set in both directions. Every page states a verification_method beside its date, so last_verified says what it actually proves instead of implying someone ran something. Contract prose carries check markers tied to a coverage matrix that separates machine-enforced rules from ones only a reviewer can judge. Scoped exemptions replace a glob that could silence anything: three rules, each with a stated reason. Component repositories can record deviations in .axelerant/adr/, which they previously had nowhere to put. The isolated verifier no longer runs documented commands on the host. Source is piped into a container with no mount, no network, dropped capabilities, a non-root user and resource limits, from an image named by digest, and it refuses to fall back. The coupling reminder reads the GitHub API instead of executing pull-request code with a write token, and skips forks. The sample is now an executable service: an HTTP API over a SQLite queue with leases, bounded retries and a local receipt adapter, covered by nine tests including concurrent claims and stale workers. Its runbooks are verified by isolated drills that the pages label as simulations. ADRs 0001 and 0002 are kept and superseded rather than rewritten; 0003 records the local-runtime boundary. The Mermaid heuristics that rejected every semicolon and counted end across diagram languages are gone. They produced false errors on valid diagrams; rendering is what establishes parse success. What I changed before taking it. The lockfile is regenerated rather than left behind its manifest, so npm ci resolves instead of failing and taking the diagram gate down with it. The workflow template is a real file an adopter can copy, pinned to this release, with the renderer kept for any other reviewed commit; the candidate had replaced it with a stub that fails on purpose, which contradicts every instruction that says to copy it. render_workflow.py now emits the coupling job and its permissions, which it had dropped. Config validation reports a missing field once instead of also calling it invalid. The contract no longer carries its own release status; that belongs in the changelog. Verified here: 40 contract tests, 9 sample tests, both audits clean, plugin, workflow and coverage checks, npm ci against the regenerated lockfile, npm audit at zero, and all five diagrams rendered with the pinned renderer. Hosted CI, the external-link pass and a real Claude plugin load remain unverified. Co-Authored-By: Claude Opus 5 (1M context) Claude-Session: https://claude.ai/code/session_01XmzqahsmAXBn5wJab92jP2