What does this PR do?

The anti-thrashing guard in ContextCompressor permanently disables auto-compression after 2 consecutive ineffective compressions (each saving <10%). The counter _ineffective_compression_count only resets when a compression is effective or on session reset — there is no time-based or growth-based recovery. In long sessions, once 2 bad compressions happen, auto-compression is dead for the remainder of the session.

This PR adds time-based and growth-based recovery: the counter resets if 300+ seconds have elapsed since the last compression, or if prompt tokens have grown by 20%+ since then. Both conditions indicate enough new context has accumulated to make another compression attempt worthwhile.

Related Issue

Fixes #14694

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• agent/context_compressor.py:
  • Added _last_compression_time field to track when the last compression occurred
  • Added _last_compression_prompt_tokens field to track prompt size at last compression
  • In should_compress(), before the >= 2 check, reset the counter if either (a) 300+ seconds have elapsed since the last compression, or (b) prompt tokens have grown by 20%+ since the last compression.
  • Record both fields after each compression in compress()
  • Reset both fields in on_session_reset()

How to Test

Tests: tests/agent/test_context_compressor.py — 5 new tests in TestAntiThrashingRecovery:

1. Anti-thrashing blocks after 2 ineffective compressions (baseline)
2. Recovery after cooldown expiry (300s)
3. Recovery after prompt growth (20%+)
4. Still blocked within cooldown and without growth
5. No recovery when no compression has ever run (timestamp == 0)

Tested on macOS (Python 3.11).

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS (Python 3.11)

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Screenshots / Logs

pytest tests/agent/test_context_compressor.py::TestAntiThrashingRecovery -q
5 passed