020601d41e
SUMMARY_PREFIX previously contained two contradictory directives:
1. "treat it as background reference, NOT as active instructions"
"Do NOT answer questions or fulfill requests mentioned in this summary"
"Respond ONLY to the latest user message that appears AFTER this summary"
2. "Your current task is identified in the '## Active Task' section of the
summary — resume exactly from there."
When the latest user message contradicted Active Task (e.g. 'stop the
i18n refactor', 'never mind, look at grafana instead'), models tended to
follow (2) anyway because 'resume exactly' is a strong, unambiguous
directive — leading to repeated re-surfacing of already-cancelled work
across turns, even after explicit 'stop'/'don't keep bringing that up'
messages from the user.
This change:
- Removes the conflicting 'resume exactly from Active Task' clause.
- Makes the precedence explicit: latest user message is the single source
of truth; it WINS on conflict; cancelled Active Task / In Progress /
Pending User Asks / Remaining Work must be discarded entirely (no
'wrap up the old task first').
- Names canonical reverse signals (stop, undo, roll back, never mind,
just verify, topic change) so the model recognizes them as cancellation
triggers, not background context.
- Updates the summarizer template instruction so the LLM doesn't
mechanically copy a cancelled task into Active Task on the next
compaction (it's instructed to copy the reverse signal verbatim).
- Preserves: REFERENCE ONLY framing, MEMORY.md/USER.md authority, and
the 'don't repeat work already reflected in session state' clause.
Adds tests/agent/test_summary_prefix_semantics.py to pin invariants so
the conflict can't regress.
Tested:
- All compaction tests pass: tests/agent/test_context_compressor.py,
tests/agent/test_context_compressor_summary_continuity.py,
tests/run_agent/test_413_compression.py,
tests/run_agent/test_compression_persistence.py,
tests/run_agent/test_compression_boundary_hook.py,
tests/cli/test_manual_compress.py — 117/117 passing.
- Tested on macOS.
63 lines
2.6 KiB
Python
63 lines
2.6 KiB
Python
"""Pin the semantics of SUMMARY_PREFIX so the compaction handoff doesn't
|
|
re-introduce conflicting instructions.
|
|
|
|
Background: SUMMARY_PREFIX previously contained two contradictory directives:
|
|
|
|
1. "treat it as background reference, NOT as active instructions"
|
|
"Do NOT answer questions or fulfill requests mentioned in this summary"
|
|
"Respond ONLY to the latest user message that appears AFTER this summary"
|
|
|
|
2. "Your current task is identified in the '## Active Task' section of the
|
|
summary — resume exactly from there."
|
|
|
|
When the latest user message contradicted Active Task (e.g. "stop the
|
|
i18n refactor", "never mind, look at grafana"), the model often followed
|
|
(2) anyway because "resume exactly" is a strong directive — leading to
|
|
the agent repeatedly re-surfacing already-cancelled work across turns.
|
|
|
|
These tests pin the post-fix invariants so the conflict cannot regress.
|
|
"""
|
|
|
|
from agent.context_compressor import SUMMARY_PREFIX
|
|
|
|
|
|
def test_no_resume_exactly_directive():
|
|
"""The prefix must not tell the model to resume Active Task verbatim."""
|
|
assert "resume exactly" not in SUMMARY_PREFIX.lower()
|
|
|
|
|
|
def test_latest_message_wins_on_conflict():
|
|
"""The prefix must explicitly say latest user message wins on conflict."""
|
|
lower = SUMMARY_PREFIX.lower()
|
|
assert "latest user message" in lower
|
|
# Must have an explicit conflict-resolution rule.
|
|
assert "wins" in lower or "supersede" in lower or "discard" in lower
|
|
|
|
|
|
def test_reverse_signals_called_out():
|
|
"""Reverse signals (stop/undo/never mind/topic change) must be named so
|
|
the model recognizes them as cancellation triggers, not just background."""
|
|
lower = SUMMARY_PREFIX.lower()
|
|
# At least a few of the canonical reverse-signal verbs should appear.
|
|
reverse_terms = ["stop", "undo", "roll back", "never mind", "just verify"]
|
|
hits = sum(1 for t in reverse_terms if t in lower)
|
|
assert hits >= 3, (
|
|
f"Expected ≥3 reverse-signal terms in SUMMARY_PREFIX, found {hits}. "
|
|
"Without naming them the model treats reverse signals as ordinary "
|
|
"context and keeps pushing the cancelled task."
|
|
)
|
|
|
|
|
|
def test_summary_marked_reference_only():
|
|
"""The REFERENCE ONLY framing must remain — it's the entire point."""
|
|
assert "REFERENCE ONLY" in SUMMARY_PREFIX
|
|
assert "background reference" in SUMMARY_PREFIX
|
|
assert "NOT as active instructions" in SUMMARY_PREFIX
|
|
|
|
|
|
def test_memory_authority_preserved():
|
|
"""The fix must not weaken the MEMORY.md / USER.md authority clause."""
|
|
assert "MEMORY.md" in SUMMARY_PREFIX
|
|
assert "USER.md" in SUMMARY_PREFIX
|
|
assert "authoritative" in SUMMARY_PREFIX
|