Merged
Conversation
Three composable slash commands extracted from session methodology: /anatomize <target> — Full two-phase pipeline: Phase 1: Document all instances of target with identity, architecture, dependencies, constraints, and limitations in structured markdown. Phase 2: Launch 6 parallel adversarial subagents with strict rubrics (convex wins, extending capabilities, reducing steps, improvement integration, accuracy fidelity, structural quality), synthesize findings, and apply improvements targeting 4+/5 across all dimensions. /adversarial-review <artifact> — Standalone Phase 2: 6-dimension parallel adversarial evaluation with scoring, findings, recommendations, and automatic application of improvements. /extract-anatomy <target> — Standalone Phase 1: Comprehensive anatomical extraction with identity, architecture, dependencies, constraints per item. Outputs structured markdown with glossary, quick reference matrix, and cross-references. https://claude.ai/code/session_01CJnBiMXiuhsnvvvcqAMPYS
New slash commands extracted from session methodology: /fan-out-review <target> — The parallel adversarial subagent pattern formalized as a reusable skill. Documents the fan-out → synthesize → apply pipeline with rubric templates for 4 artifact types: documentation, code, tests, and architecture. Includes custom rubric design guide for anything that doesn't fit the templates. /considerations <context> — The 7 key considerations as pre-flight checklist and in-flight guard-rails: 1. Adversarial eval most valuable on first drafts (diminishing returns) 2. Parallel evaluation beats sequential (narrow mandates, no groupthink) 3. Fan-out → synthesize → apply is the core reusable unit 4. Scoring forces accountability (no hand-waving) 5. Source-of-truth extraction catches documentation lies 6. Prioritize convex easy wins (impact/effort ratio) 7. Know when to stop (4+/5 or <3 findings per round) Includes calibration table, priority formula, and stopping criteria. /review-docs <file> — Focused documentation review with 6 parallel rubrics tuned for reference material. /review-code <file> — Focused code review with 6 parallel rubrics tuned for implementation (correctness prioritized over cleanup). /audit-completeness <artifact> — Cross-reference any artifact against its source of truth. Produces structured delta report with severity classification. Includes "common documentation lies" checklist. https://claude.ai/code/session_01CJnBiMXiuhsnvvvcqAMPYS
Both /extract-anatomy and /anatomize now use parallel subagent orchestration for the extraction phase, not just the evaluation phase. Phase 1 now fans out to 6 simultaneous extraction subagents: E1: Identity & Inventory (master checklist) E2: Inputs, Outputs & Schemas (parameter tables from source) E3: Internal Mechanics & State Effects (runtime behavior) E4: Dependencies & Relationships (dependency graph, peer map) E5: Constraints, Limitations & Availability (hard/soft limits) E6: Configuration & Environment (env vars, CLI flags, config files) Key additions: - Phase 0 (Reconnaissance): fast discovery scan before extraction - Phase 1.5 (Merge): conflict resolution rules (schema > observed) - Validation checklist before proceeding to evaluation - Scaling strategy: split E2/E3 by category for >20 instances - Optional E7 (cross-system) and E8 (versioning) subagents - Rationale table: why parallel extraction beats serial Full pipeline is now: Recon (1) → Extract (6‖) → Merge → Evaluate (6‖) → Synthesize → Apply https://claude.ai/code/session_01CJnBiMXiuhsnvvvcqAMPYS
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.
No description provided.