compacting-context

name	compacting-context
description	Context compression for verbose outputs. Use when conversation context exceeds 70% capacity or when summarizing large outputs.

Context Compaction

Compresses verbose outputs into concise summaries while preserving essential information. Use proactively to maintain context efficiency.

Trigger Conditions

• Context utilization exceeds 70%
• Output exceeds 500 lines
• Repeating information from earlier in conversation
• Full file contents included when excerpts suffice

Workflow

1. Identify verbose content - Large outputs, repeated info, full files
2. Extract essential information - Key facts, decisions, action items
3. Apply compression strategy - Summarize, reference, abbreviate
4. Verify completeness - Ensure no critical information lost
5. Output compact summary - Concise format with references

Compression Strategies

Reference Instead of Repeat:

# Before
The function in src/auth.py handles authentication by...
[50 lines of explanation]

# After
See `src/auth.py:42-85` for auth implementation details.

Summarize Lists:

# Before
Test 1 passed, Test 2 passed, Test 3 passed... [15 more]

# After
All 18 tests pass. Coverage: 87%.

Table Format:

| File | Status | Lines Changed |
|------|--------|---------------|
| auth.py | Modified | +15, -3 |
| tests/test_auth.py | Added | +45 |

Output Format

Status Updates: Single line with metrics Code Changes: File:line references, not full content Test Results: Summary counts, only show failures in detail Errors: Line reference + brief description

Quality Standards

• No critical information lost
• References are accurate (file:line)
• Summary is actionable
• Format is consistent