Add fix analysis to reviewer and fix_gate_mode#78
Merged
Conversation
The reviewer now recommends how to fix each issue (FIX, ALTERNATIVE, CONFIDENCE) rather than only diagnosing what's wrong. This gives the user visibility into the system's fix strategy before the executor re-attempts, building confidence toward eventually removing human gates. A new fix_gate_mode (gated/auto) mirrors task_gate_mode, allowing users to auto-approve fix analyses independently. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When fix_gate_mode is auto, a counter tracks review-failure cycles per task. After 3 failed attempts, the loop escalates to the user regardless of fix_gate_mode. Counter resets on each new task. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Step 0 now checks branch and diff state to auto-detect which flowchart keys need auditing, reducing unnecessary full audits. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The tripwire counter is now stored in the implementation tracking file rather than held in memory. Resets to 0 on new task (stage A) and on session resume. Context refresh recovery checks for mid-loop state via fix_attempts > 0. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Auto-proceed is the narrow exception (auto + under tripwire threshold), checked first. All other cases — gated mode or tripwire triggered — fall through to user presentation. Removes backwards "see below" reference and "follow gated path above" routing. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace descriptive titles with clear conditional h4 headings that mirror the structure in section D (Task Gate). Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
leeovery
added a commit
that referenced
this pull request
Feb 6, 2026
…itions) V3's integration context created a "convention gravity well" where early design choices got documented as established patterns and later tasks faithfully propagated them. V2 without this mechanism won 21/23 tasks. Removes: integration context file accumulation, plan file access for executor, expanded exploration instructions, codebase cohesion review dimension, Convention Consistency quality section. Keeps: fix recommendations (PR #78), fix_gate_mode (PR #78), full re-attempt context (PR #77), polish agent (PR #80). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
6 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
1 participant
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
needs-changesissue includesFIX(recommended approach),ALTERNATIVE(optional, when multiple valid approaches exist), andCONFIDENCE(high/medium/low) — giving visibility into the system's fix strategy before the executor re-attemptsfix_gate_mode: Mirrorstask_gate_mode(gated/auto) but controls the review-changes gate independently. Users can auto-approve fix analyses while keeping the task approval gate, or vice versafix_gate_mode: gatedadded to template, reset on session resume, checked during context refresh recoveryFiles changed
agents/implementation-task-reviewer.mdskills/technical-implementation/references/steps/invoke-reviewer.mdskills/technical-implementation/references/steps/task-loop.mdfix_gate_mode, auto option addedskills/technical-implementation/SKILL.mdContext
When the reviewer returns
needs-changes, it currently says what's wrong but not how to fix it. The executor then independently decides the approach with no user visibility. This makes it hard to build confidence toward removing human-in-the-loop gates.With fix analysis, the user can see the system's reasoning and planned approach before re-execution. Over time, consistently good fix recommendations build the trust needed to enable
fix_gate_mode: auto.Test plan
needs-changes— verify reviewer output includes FIX/CONFIDENCE per issuea/autoat the fix gate — verifyfix_gate_modeupdates toautoin tracking filefix_gate_modetogatedfix_gate_mode: autoskips the fix gate and re-invokes executor directly🤖 Generated with Claude Code