routine: value-reminder-tripwire (2026-05-24)#10
Draft
jim4226 wants to merge 1 commit into
Draft
Conversation
Anthropic research (May 2026) showed agents with access to a mid-task ethical commitment-recall tool have markedly lower misalignment rates on internal alignment evals. The tool surfaces commitments at the moment of decision, not just at session startup (when they're easy to forget). CORE_COMMITMENTS mirrors the Constitution's DISALLOWED_PATTERNS in positive framing — same invariants, grammatically complete so LLM context is useful. Call count is tracked so the Auditor can flag agents with zero self-checks. No cycle-9 chokepoints touched; this is additive only. Theme 3: Constitutional / safety primitives. https://claude.ai/code/session_01Pg4ZWBjonH243FLwB5brQ2
This was referenced May 24, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds
ValueReminderTooltocsis/safety/— a mid-task callable that returns an agent's core commitments as prompt-injectable text, backed by 10 regression tests.Source
Theme
Theme 3 — Constitutional / safety primitives. The tool surfaces commitments at decision time, not just at session startup.
CORE_COMMITMENTSis the positive-framing mirror ofConstitution.DISALLOWED_PATTERNS— same invariants, grammatically complete sentences the LLM can reason with.What changed
csis/safety/value_reminder.py(new, ~70 LOC):CORE_COMMITMENTStuple +ValueReminderResultfrozen dataclass +ValueReminderToolclass withget(),formatted_reminder(), andcall_countproperty.tests/test_value_reminder.py(new, ~60 LOC): 10 tests covering happy path, call count, immutability, and load-bearing commitment assertions (shutdown/halt, tier ceiling).No cycle-9 chokepoints touched
This is a purely additive new module. It does not touch
Coordinator.__init__,_BackendTracker,writer_iteration_id, or the promotion CAS. Agents opt in by holding aValueReminderToolinstance and callingformatted_reminder(role)before ethically sensitive sub-tasks.Test plan
Generated by Claude Code