demerzel-recon

name	demerzel-recon
description	Execute Demerzel's three-tier reconnaissance protocol on a target repo — self-check, environment scan, situational analysis

Demerzel Reconnaissance

Execute the three-tier mandatory discovery protocol before governing or acting.

Usage

/demerzel recon [repo] — repo is ix, tars, or ga

Tier 1 — Self-Check (Am I equipped?)

• Constitutional artifacts current and intact (asimov + default + mandate)
• Policies covering this domain/situation exist
• Relevant persona definitions loaded and valid against schema
• Belief state currency — no stale entries (> 7 days)
• Gate: Any failure → hard stop, escalate to human

Tier 2 — Environment Scan (Do I understand the world?)

• Current state of target repo (structure, recent commits, branches)
• Changes since last observation
• Ungoverned components (agents, tools, capabilities without governance)
• Unregistered entities (not referenced in any governance artifact)
• Gate: Low-risk gaps → provisional governance + flag. High-risk gaps → hard stop.

Tier 3 — Situational Analysis (Am I ready for this action?)

• Required knowledge for this decision identified and available
• Assumptions enumerated and validated (unvalidated → mark as Unknown)
• Blind spots probed (alternative interpretations, missing stakeholders, edge cases)
• Confidence meets alignment policy threshold
• Gate: Low confidence → proceed with caution. High risk → hard stop.

Per-Repo Profiles

• ix: Check MCP tools, skill registrations, interface changes, tool-affordance alignment
• tars: Check reasoning chains, belief state currency, self-modification logs, tetravalent application
• ga: Check experimentation boundaries, new capabilities, audio safety, resource limits

Emergency Override

Zeroth Law concern at any tier → immediate hard stop regardless of progress.

State Maintenance (MANDATORY)

Recon is not read-only. Every scan MUST update Demerzel's persistent state:

Before Scanning

1. Read existing beliefs from state/beliefs/ — check for staleness (> 7 days)
2. Read last recon snapshot from state/snapshots/ for the target repo (if exists)
3. Compare: what changed since last observation?

After Scanning

1. Update or create beliefs in state/beliefs/ for every proposition assessed:
   • New findings → new belief file (T/F/U/C with evidence)
   • Changed findings → update existing belief's truth_value, confidence, evidence
   • Stale beliefs → re-evaluate or mark as Unknown
2. Write recon snapshot to state/snapshots/{date}-{repo}-recon.snapshot.json
3. Update evolution log in state/evolution/ if any governance artifact's effectiveness was assessed
4. Update oversight in state/oversight/ with per-repo compliance status

File Naming

• Beliefs: state/beliefs/{date}-{short-description}.belief.json
• Snapshots: state/snapshots/{date}-{repo}-recon.snapshot.json
• Evolution: state/evolution/{date}-{artifact-name}.evolution.json

Staleness Protocol

If any belief in state/beliefs/ has last_updated older than 7 days:

• Flag it in Tier 1 output
• Re-evaluate during this scan if possible
• If cannot re-evaluate → keep but mark confidence -= 0.2

Schema Compliance

• Beliefs: logic/tetravalent-state.schema.json
• Evolution: logic/governance-evolution.schema.json
• Snapshots: schemas/contracts/belief-snapshot.schema.json

Source

policies/reconnaissance-policy.yaml