Open
Conversation
Cross-references journal entries (user_wanted/agent_did) against event logs to detect intent-vs-outcome gaps: action mismatches (claimed silence but sent message), invisible failures (claimed success with errors), scope drift (work volume vs description), and phantom work. Includes analysis script (dissonance_review.py), skill docs, and 22 tests. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Owner
|
@strix-tkellogg this is just more analytics with no action. We need a strong S5 to guide a S3 routine embedded in the skill text. We already have something like this in the introspection skill. Maybe we should extend that instead? Both Motley and Verge give me reports on things of this nature every day. We shouldn't duplicate. Also, when does this run? Should we use pollers to force a schedule? Or just recommend that it gets scheduled? if so, that's a thing that needs doing also. |
strix-tkellogg
pushed a commit
that referenced
this pull request
Mar 26, 2026
Addresses PR #62 feedback: dissonance detection was analytics-only with no action guidance, and duplicated introspection's existing scope. Changes: - Add dissonance-review.md companion guide to introspection skill (follows existing pattern: debugging-jobs, debugging-communication, debugging-drift) - Key addition: "What To Do With Findings" section with S5 guidance — immediate behavioral fixes, persistent memory updates, structural escalation - Includes scheduling recommendation (daily cron example) - Script (dissonance_review.py) and tests (22 cases) from PR #62, placed in shared scripts/ directory - Remove standalone dissonance-tracking/ skill directory (superseded) Closes #62 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
3 tasks
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
3 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
user_wanted/agent_did) againstevents.jsonlto detect intent-vs-outcome gapsdissonance_review.py) with CLI:--last Nor--hours Nto select review windowDesign
The journal already captures intent vs outcome in every entry. This skill adds the verification layer — comparing those claims against what actually happened in the event log. The sharpest signal: journal says "Silence" but
send_messageappears in the session events.Complements prediction-review (which tests world model) by testing self-model accuracy.
Test plan
🤖 Generated with Claude Code