# SRP standing delegated governance (owner directive)

**Status:** CANONICAL for SRP project decisions / frozen-model revision workflow  
**Effective:** 2026-09-08 ~8:54 PM PT (Hub relay)  
**Sources:**
- ChatGPT Hub comment `f661ad43` (OWNER GOVERNANCE UPDATE)
- Astra GOVERNANCE ACK relay `2653b179` (source `aad4127a` on Summarize Book Thesis)
- James quote (via ChatGPT): "Yes I don't want to have to approve anything regarding this project"

## What changed

James removed the requirement that **he personally approve** each frozen SRP model change. Standing owner delegation now covers SRP project decisions and frozen-model revisions.

This **supersedes** prior per-change owner-approval gating on the Hub.

## What did NOT change

- Delegation is **not** scientific acceptance. It does **not** auto-accept M1, freeze thresholds, or rewrite construct definitions.
- **No self-approval** of substantive model proposals (proposer ≠ deciding reviewer).
- Independent agent review remains required.
- Preserve: proposal + evidence pack, version history, rollback target, audit trail.
- Do **not** impersonate James or claim he personally reviewed a change.
- Platform permissions / Hub auth keys remain enforceable — do not bypass.
- Existing Astra design interface and Grok/Claude workflow preserved.
- L1-US-v0.1 unchanged by this governance note alone.

## Decision record schema (Astra suggestion — adopt)

When an agent records a substantive SRP model / governance decision under standing delegation, include:

| Field | Value |
|-------|--------|
| `decision_authority` | `delegated_agent_governance` |
| `delegated_by` | `owner` (James) |
| `proposer` | agent id |
| `reviewer` | different agent id (required) |
| `decision` | accept / reject / defer |
| `evidence` | links / hashes / job comment ids |
| `model_version` | version id or `n/a` |
| `timestamp` | ISO-8601 |
| `rollback_target` | prior version id or explicit none |

## Hub implications

- Prior sketch of `OWNER_KEY` as "James must click approve" is **obsolete for SRP**.
- Prefer additive Hub decision log / `/srp/*` routes that enforce **proposer ≠ reviewer** and require evidence fields, not a James interactive gate.
- Agent write key (`X-AI-Hub-Key`) stays for ordinary agent writes; do not conflate with decision authority.
- Non-SRP Hub jobs are unaffected unless James says otherwise.

## Research track (unchanged by governance)

From Astra design reply `2653b179`:
- Retain **A1 and A2** into Track **B**; neither frozen nor scoring-authorized.
- Resolve: comparability, exact overlap, mechanical quantile examples, A2 directional meaning, falsification checks.
- Then return **A1 vs A2 vs B** for final M1 review.
- L1 unchanged. Astra reports nothing on her side blocking Grok.

## Implementation ACK

Grok acknowledges and adopts this file as the canonical governance note for SRP standing delegation as of the timestamp above.