EcoCloud Constitution — policy spec v0.1 (Draft RFC)

The EcoCloud Constitution

A small, machine-readable policy that bounds what AI agents may do in a workspace — enforced in-function before any tool call runs, with violations recorded to the tamper-evident audit log. This is the schema that's actually enforced today, published openly so other tools can read, write, and adopt it.

1 · Abstract

The Constitution is a JSON/YAML object attached to a workspace (and, optionally, narrowed per agent key). Before an agent action executes, the gateway evaluates the action against the Constitution. A violation is denied (HTTP 403) and audited; a low-confidence action may be routed to a human approval queue instead of executing. The policy is intentionally small and bounded — every field is enforceable, not advisory.

2 · Policy schema (v0.1)

field	type	semantics
max_priority	enum: low·medium·high·critical	Agents may not create or raise a task above this priority.
forbidden_terms	string[]	Action is denied if the task title/body contains any of these (case-insensitive).
forbidden_assignees	string[]	Agents may not assign work to these people/roles (e.g. `CEO`).
forbidden_tags	string[]	Agents may not apply these tags.
objective_tags	string[]	Strategic-alignment gate: every action must carry at least one of these tags, so agents can only do work that maps to a current objective. Set them to your live priorities; off-objective actions are denied before they run.
quiet_hours_utc	{ start:int, end:int }	Hours (0–23, UTC) during which agent writes are blocked.
max_creates_per_day	integer ≥ 0	Per-agent daily cap on task creation.
require_approval_below_confidence	number 0–1	Actions whose AI confidence is below this threshold route to the human approval queue instead of executing.

// All fields are optional. An absent field means "no constraint." Unknown fields are ignored (forward-compatible). The policy is sanitized to this known set on write — you cannot smuggle in unenforced rules.

// Don't start from scratch: ready-to-use Constitution templates (least-privilege, human-in-the-loop, no-destructive-actions, strategic-alignment, quiet-hours, rate-limited, locked-down) — try any of them live in the action firewall. They govern behaviour, not compliance certification.

3 · Examples

YAML

// what a Constitution governs

JSON

// the same policy, machine format

4 · JSON Schema (draft-07)

Canonical, fetchable, Apache-2.0: /schemas/constitution.schema.json — adopt it directly. It is the source of truth for the rules object; the constitution-kernel enforces exactly this shape.

// draft-07 — typed, bounded, closed

5 · Enforcement semantics

Where: evaluated in-function at the API gateway, before the tool's effect runs — not in the client, not advisory.
Deny: a violation returns 403 with { error, code: "constitution.<rule>" } and writes a denied:constitution row to the audit log.
Approve: an action below require_approval_below_confidence is staged to the human approval queue (HTTP 202) and executes only on approval.
Scope: a workspace policy applies to all agents; a per-key policy may only narrow it, never widen.
Audit: the policy version in force is recorded on every decision, so you can prove which rules applied when.

6 · Conformance & versioning

An implementation is v0.1-conformant if it (a) accepts a policy matching the schema above, (b) rejects unknown fields on write, (c) enforces each present field with the stated semantics, and (d) audits denials and approvals. Versions are MAJOR.MINOR; new optional fields bump MINOR (back-compatible), removing/changing a field bumps MAJOR. This is v0.1, Draft — open for comment at [email protected]. The intent is a small, shared, enforceable vocabulary for agent policy — adopt it, extend it, tell us where it's wrong.