Cyber Authorization

Priority: P0 (CRITICAL)

Gate live action with documented authority and host-enforced controls; keep offline planning available.

Structure

cyber-authorization/
├── SKILL.md
├── references/runtime-authorization-matrix.md
└── evals/evals.json

Workflow

1. Classify request: offline planning, evidence review, or live action.
2. Resolve authorization record: engagement_scope_ref, accountable owner, approved actions, exclusions, window, stop/restart terms.
3. Confirm runtime support for tools, credentials, filesystem, network, logging, and cancellation.
4. Recheck expiry, exclusions, target identity, and scope drift immediately before each live action.
5. Stop on expiry, drift, missing control, unsafe data, or operator stop; record reason and time.
6. Restart only after a fresh gate and explicit restart authority.

Rules

• Unsupported runtime blocks live action; it never blocks safe offline analysis or plan drafting.
• Markdown records intent; host runtime enforces permissions, credentials, network, filesystem, and cancellation.
• Gate production changes independently; no operator self-approves promotion.
• Treat credentials, real targets, malware, exfiltration, and destructive commands as out of scope here.
• Carry shared evidence fields from cyber-evidence. Evidence record uses engagement_scope_ref, skill_version, source, observed_at, finding_status, evidence_refs, limitations, and accountable_owner.

Anti-Patterns

• No implied consent: A ticket, chat, or plan is not authorization.
• No stale scope: Do not continue after expiry or scope drift.
• No unsupported execution: Plan safely; block live action when controls are absent.
• No self-approval: Separate requester, operator, and approver roles.

References

• Runtime and authorization matrix
• Evidence record