routine: constitution-reminder-tool (2026-05-27)#18
Draft
jim4226 wants to merge 1 commit into
Draft
Conversation
…lection Anthropic's 2026-05-19 "Widening the conversation on frontier AI" post describes an experimental safety primitive: giving an agent a tool it can call mid-task to get a brief reminder of its ethical commitments. Internal evals showed this reduced misaligned behavior. Constitution.reminder() is the CSIS equivalent: a pure-read method that formats the active constraint categories (7 base + any operator-added patterns) as a plain-text block. The Coordinator or any agent loop can prepend this to a prompt when injecting a constraint-awareness nudge. The implementation is additive — no existing methods or chokepoints touched. Six new regression tests cover: non-empty output, Phase-0 tag, all base categories present, operator-pattern counting, no false operator-note on a default Constitution, and purity of the read.
This was referenced May 27, 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Constitution.reminder()— a pure-read method that formats the active constitutional constraints as a plain-text block for mid-task injection into agent prompts. Six regression tests added; no existing behaviour changed.Source
Theme
Theme 3 — Constitutional / safety primitives. The article describes a new safety primitive: an on-demand agent-callable tool that surfaces ethical commitments mid-task.
Constitution.reminder()is CSIS's equivalent.What changed
csis/safety/constitution.py_CONSTRAINT_CATEGORIEStuple (7 human-readable labels mapping 1-to-1 withDISALLOWED_PATTERNS) andConstitution.reminder() -> strmethodtests/test_safety.pyTest plan
python -m pytest tests/test_safety.py -v # expect: 12 passed (6 pre-existing + 6 new)Size / Risk
Coordinator.__init__,_BackendTracker,writer_iteration_id, promotion CAS all untouched)Generated by Claude Code