✨ Add verification agents and parallel task execution - #4
Merged
Conversation
… research Codebase research beyond the agents-context README can now dispatch up to 3 read-only Explore subagents in a single parallel batch (one per domain area) instead of a single agent. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Research subagents now report file-contention hot spots, and grouping uses them: slices that fight over the same files get merged or an honest dependency instead of fake parallelism - tasks.md gains an Execution Waves subsection under Execution Order — waves of groups with no mutual dependencies and disjoint file sets that step3 can dispatch concurrently after validating against plans - Add a self-check phase before presenting tasks.md and a scope-boundary line (this skill produces tasks.md only, never code or plans) - Template and filled-in example updated to match Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Restructure the 273-line SKILL.md into an orchestrator (48 lines) with steps/ and shared/ files — order lives only in SKILL.md; step files are topic-named and position-agnostic. plans/group-N.md gets a real template.md instead of prose inside a prompt. New verification agents, running in every mode: - implementation-reviewer + test-verifier: a parallel read-only pair checking each group before commit (diff vs. plan/spec/standards; tests meaningful, edge cases covered, nothing weakened). Orchestrated execution drives a bounded fix cycle (max 2 rounds); L mode surfaces findings at the user's review gate - adversarial-thinker: challenges the plan batch before code is written and attacks the finished feature before delivery; only findings with a concrete failure scenario count Parallel execution is now scheduled, not ad hoc: pre-planning validates step2's Execution Waves against the plans' file operations and dispatches executors wave-by-wave. Tests: step3 content assertions became directory-recursive; 15 new assertions cover the structured shape, the agents, read-only and bounded-fix-cycle constraints, and waves. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Sync workflow.md and implementation.md with the step2/step3 changes: verification agents (implementation-reviewer, test-verifier, adversarial-thinker), wave-based parallel execution, and the adversarial delivery review gate before the full-suite backstop. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Installed users only receive updates when plugin.json's version changes. Co-Authored-By: Claude Fable 5 <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
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.
✨ What's New
🧪 Test Plan
🤖 Generated with Claude Code