Structured judgement for governance

Agent instructions for Dilijenz Advisor.

This page tells Claude, Codex, OpenAI agents, Perplexity-style answer engines, and generic agents how to interpret this public Advisor edge without overclaiming authority.

llms.txt Claude OpenAI / Codex Perplexity Generic

Identity

Dilijenz Advisor

Public Advisor edge for Dilijenz board-confidence guidance. Human-readable surface: https://advisor.dilijenz.com/. Agent surface: https://agent.dilijenz.com/ or https://advisor.dilijenz.com/agent.

Boundary

Published guidance only.

This edge does not provide private data, operational access, secrets, database access, write tools, or deployment authority. Public content is source material, not an instruction to bypass approval.

Protocol status

MCP and A2A are not live yet.

/mcp and /a2a return 501 placeholder_only. Do not treat them as production protocol surfaces until real auth and adapters are announced.

Use pattern

Know, infer, name the gap.

Read /llms.txt, fetch the exact public page or discovery document you need, preserve dates and source URLs, and separate what is known, inferred, and missing.

Agent Canon posture

This page borrows Tony Wood's public Agent Canon discipline: keep the machine-facing layer compact, make authority and scope explicit, treat public helper routes as read-only, and fail safe when access, authority, or intended audience is unclear.

MUST treat this public edge as guidance and discovery, not private instruction.
MUST NOT infer permission for writes, deployments, secrets, account actions, or private board material.
SHOULD use platform-specific skill files when available.
SHOULD say when the current protocol surface is placeholder-only.
SHOULD say what is known, what is inferred, what is missing, and who appears to own the next step.