Cino AgentOps
StarNet-native · v1.0.0

Agents that know what to do next.

Cino AgentOps is a compact coordination layer for StarNet: scope the evidence, resolve authority, route work, navigate code, choose the smallest useful test, and hand back proof without blurring the boundaries.

Install in StarNet Inspect the exact v1 source
Install

Discover it from the site.

Path-aware StarNet builds can discover the registry from this page URL. The direct registry URL is kept as a transparent fallback for older builds and for manual inspection.

Automatic StarNet discovery

On a path-aware StarNet build, use the public site URL in Skill Exchange discovery.

https://cinii05.github.io/cino-toolkit-public/agentops/
Open Skill Exchange → Discover.

Use discovery rather than installing unknown bytes directly.

Paste the site URL above.

StarNet resolves the page's well-known registry and returns the Cino AgentOps entries.

Inspect before installing.

StarNet freezes the fetched bytes, calculates the canonical package digest, runs the skill guard, then installs only the inspected artifact through its normal approval path.

No silent authority expansion. Installing AgentOps does not grant deployment, account, network, repository, or production permissions that the agent did not already have.
Core pack

Small skills. Clear ownership.

Each procedure owns one coordination job. Specialist project skills still take precedence when they fully own a workflow.

cino-context-funnel

Narrows broad, multi-source work before authority resolution starts.

cino-source-of-truth

Resolves an explicit authority conflict without pretending every source is equally current.

cino-agent-router

Routes a bounded delegation decision without hijacking a larger implementation workflow.

cino-repo-navigator

Maps the smallest relevant code surface before editing.

cino-test-router

Selects the smallest sufficient verification instead of defaulting to every test.

cino-evidence-handoff

Returns a compact completion receipt only at a genuine handoff boundary.

cino-plan-before-build

Coordinates non-trivial implementation while keeping plan, edit, commit, push, merge and deploy authority separate.

cino-skill-governor

Owns reusable workflow classification and the skill lifecycle, including promotion and rollback.

Internal reference: cino-skill-governor-approval-model is published for integrity and dependency resolution, but is intended to be loaded only when the governor reaches a persistence decision.
Release proof

Public bytes match the accepted v1.

The public GitHub sources and exported package envelopes were re-fetched after publication and compared against the registry digests.

9 / 9public SKILL.md package digests matched
9 / 9exported envelope digests matched
A1–A12trigger and precedence acceptance completed before v1 promotion
Boundaries

Procedure, not permission.

AgentOps helps an agent decide what workflow applies and where to stop. It is deliberately not a back door around the host product's permission model.

What it can do

  • Coordinate evidence and authority checks.
  • Route bounded work.
  • Guide repository navigation and verification.
  • Require evidence before completion claims.
  • Govern the lifecycle of reusable procedures.

What it cannot grant

  • Repository write access.
  • Production deployment authority.
  • Credentials or account permissions.
  • Approval to merge, publish, purchase or message externally.
  • Trust in third-party skill bytes without inspection.