context-admission-operator

name	context-admission-operator
description	Use when command output, CI logs, ART packets, runtime diagnostics, repo context, or other operational context may be large or sensitive and should be captured, redacted, packetized, budgeted, or admitted through Context Governance Gateway instead of pasted raw.

Context Admission Operator

Use this when operational context is large, sensitive, repetitive, or likely to burn tokens if pasted raw.

Read First

• docs/context-admission-governance.md
• contracts/context-behavior.yaml
• contracts/raw-context-retirement.yaml
• context-governance-gateway/AGENTS.md
• context-governance-gateway/README.md
• workspace-root/AGENTS.md

Workflow

1. Classify the context source: terminal, CI, ART, runtime, repo, security, or operator packet.
2. Preserve full raw artifacts locally or in the approved custody path.
3. Project only model-safe/operator-safe packets into agent or operator context.
4. Redact secret-like values, private IPs, hostnames, tokens, env dumps, and uncertain sensitive material by default.
5. Prefer compact ART packet commands first for ART work; use CGG packet refs for oversized output or context that should not be raw-projected.
6. Keep WGCF, OOS, platform, and security authority boundaries intact. CGG governs context admission; it does not approve work or mutate authority stores.
7. Record packet refs, digests, receipts, and redaction decisions in evidence when the packet is used for completion or review.

Guardrails

• Do not paste raw terminal, CI, ART, Kubernetes, OpenProject, or security output into model context when packet parity exists or the material is sensitive.
• Do not use CGG as a custom LLM gateway, scanner, observability backend, object store, WGCF replacement, or ART mutation authority.
• Default uncertain sensitive material to denied raw model projection.
• If CGG is unavailable, use the documented rollback path and record the limitation instead of pretending the context was governed.