// Runtime context compression for agents approaching model context limits. Defines 3 compression tiers (full/summarized/minimal) with per-artifact templates. USE FOR: reducing artifact loading size at runtime, context budget management. DO NOT USE FOR: diagnostic context auditing (use context-optimizer), Azure infrastructure.
| name | context-shredding |
| description | Runtime context compression for agents approaching model context limits. Defines 3 compression tiers (full/summarized/minimal) with per-artifact templates. USE FOR: reducing artifact loading size at runtime, context budget management. DO NOT USE FOR: diagnostic context auditing (use context-optimizer), Azure infrastructure. |
Runtime compression system that actively reduces context when agents approach model limits. Agents check approximate context usage before loading artifact files and select the appropriate compression tier.
| Tier | Context Usage | Strategy |
|---|---|---|
full | < 60% | Load entire artifact — no compression |
summarized | 60-80% | Load key H2 sections only |
minimal | > 80% | Load decision summaries only (< 500 chars) |
Before loading any artifact file:
1. Estimate current context usage (rough: 1 token ≈ 4 chars)
2. Check model limit (Opus: 200K, GPT-5.3-Codex: 128K)
3. Calculate usage percentage
4. Select tier:
< 60% → full (no compression needed)
60-80% → summarized (key sections only)
> 80% → minimal (decision summaries only)
5. Load artifact/skill using the appropriate variant
Skills also have compression tiers (digest and minimal variants):
| Context Usage | Skill Variant | Path Pattern | Approx Tokens |
|---|---|---|---|
| < 60% | Full | SKILL.md | 400-800 |
| 60-80% | Digest | SKILL.digest.md | 150-320 |
| > 80% | Minimal | SKILL.minimal.md | 40-100 |
Available for: azure-defaults, azure-artifacts,
golden-principles, context-shredding.
| Reference | File | Content |
|---|---|---|
| Compression Templates | references/compression-templates.md | Per-artifact H2 sections per tier |
Artifact template structures, H2 compliance rules, and documentation styling for agent outputs (Steps 1-7). USE FOR: generating any agent artifact, checking H2 structure compliance. DO NOT USE FOR: Azure resource configuration (use azure-defaults), Bicep/Terraform patterns (use bicep-patterns or terraform-patterns).
Deterministic Azure Policy discovery: lists effective policy assignments at subscription scope (including MG-inherited), pulls definitions and exemptions, classifies effects, filters Defender auto-assignments, and emits the governance-constraints JSON envelope via a Python script. USE FOR: 04g-Governance Phase 1 discovery, refreshing `04-governance-constraints.json`. DO NOT USE FOR: artifact writing, architecture mapping, traffic-light rendering, challenger orchestration — those stay in the parent agent.
Authoritative reference for VS Code Copilot customization mechanisms: instructions, prompt files, custom agents, agent skills, MCP servers, hooks, and plugins. Use when deciding which customization type to use, creating new .instructions.md/.prompt.md/.agent.md/SKILL.md/mcp.json files from scratch, or debugging why a customization is not loading. DO NOT USE FOR: routine file edits where the format is already known.
Azure infrastructure defaults: regions, tags, naming (CAF), AVM-first policy, security baseline, unique suffix patterns. USE FOR: any agent generating or planning Azure resources. DO NOT USE FOR: artifact template structures (use azure-artifacts), pricing lookups (read references/pricing-guidance.md on demand).
Shared IaC deploy patterns for Bicep and Terraform deploy agents: deployment strategies, circuit breaker, known deploy issues. For preflight validation (auth, governance, stop rules), see azure-validate. USE FOR: Phased deployment, circuit breaker, deploy-specific known issues. DO NOT USE FOR: Preflight validation (use azure-validate), code generation (use azure-bicep-patterns or terraform-patterns).
Machine-readable workflow DAG for the multi-step agent pipeline. Defines node types, edge conditions, gates, and fan-out patterns. USE FOR: Orchestrator step routing, resume-from-graph, workflow validation. DO NOT USE FOR: Azure infrastructure, code generation, troubleshooting.