common-context-optimization

Files (3)

• SKILL.md
• compaction.md · references/
• masking.md · references/

SKILL.md (first lines)

## **Priority: P1 (OPTIMIZATION)**

Manage the Attention Budget. Treat context as a scarce resource.

## 1. Observation Masking (Noise Reduction)

**Problem**: Large tool outputs (logs, JSON lists) flood context and degrade reasoning.
**Solution**: Replace raw output with semantic summaries _after_ consumption.

1. **Identify**: outputs > 50 lines or > 1kb.
2. **Extract**: Read critical data points immediately.
3. **Mask**: Rewrite history to replace raw data with `[Reference: <summary_of_findings>]`.
4. **See**: `references/masking.md` for patterns.

## 2. Context Compaction (State Preservation)

**Problem**: Long conversations drift from original intent.
**Solution**: Recursive summarization that preserves _State_ over _Dialogue_.

1. **Trigger**: Every 10 turns or 8k tokens.
2. **Compact**:
    - **Keep**: User Goal, Active Task, Current Errors, Key Decisions.
    - **Drop**: Chat chit-chat, intermediate tool calls, corrected assumptions.
3. **Format**: Update `System Prompt` or `Memory File` with compacted state.
4. **See**: `references/compaction.md` for algorithms.

## 3. KV-Cache Awareness (Latency)

**Goal**: Maximize pre-fill cache hits.

- **Static Prefix**: strict ordering: System -> Tools -> RAG -> User.
- **Append-Only**: Avoid inserting into the middle of history if possible.

## References

- [Observation Masking Patterns](references/masking.md)
- [Compaction Algorithms](references/compaction.md)

## Anti-Patterns

- **No raw tool dumps**: Mask large outputs immediately after extracting data.
- **No append-only growth**: Compact every 10 turns to preserve intent over dialogue.
- **No middle insertions**: Append-only history maximizes KV cache hits.

From wildbitca/ai-resources, Apache-2.0.

More from wildbitca/ai-resources