The library

codex-debate · When code review requires adversarial consensus between multiple reviewers.

markup-repair · Fixing text markup without changing wording or structure.

hiraia-curate-images · Filtering subject clusters into drawable, single-subject image concepts.

hiraia-gen-eval-queries · Generating labeled benchmark queries for retrieval system evaluation.

loop-until-dry · Iterative bug hunting until convergence on a stable result set.

completeness-critic · Identifying gaps in initial coverage then filling them iteratively.

ts-sdk-publish · Publishing SDK releases with validation and multi-stage deployment.

verify-findings-rubric · Multi-agent orchestration for verify findings rubric workflows.

verifica-report · Auditing financial notes against transactional DB when you need claim-by-claim verification with source traceability.

scaffold-feature · Scaffolding a new feature when you want parallel file generation against a plan plus a built-in consistency review.

lets-plan-workflow · Multi-phase orchestration when phases need to be planned sequentially with approval checkpoints.

evolve-batch · Iterative batch refinement when you need multiple feedback cycles with controlled evolution.

agi-cli-migration-ux · CLI migrations when you need designed UX with clear deprecation paths and fallback.

test-gap-finder · Test improvement when you need to identify which code paths are untested and which tests are weak.

codebase-survey · Onboarding or refactoring when you want a comprehensive structure survey without running code.

understand · Deep dives when you need a thorough mental model of a complex system from an agent.

adversarial-verify · Decision-making when you want an agent to systematically argue against a claim before committing.

routed_repo_review · Conducting structured code/design reviews with multiple perspectives in parallel.

finalize-testing-rules-audit · Conducting structured code/design reviews with multiple perspectives in parallel.

forge-conventions-extract · Extracting and indexing knowledge from distributed sources.

columbia-churn-2024-2026 · Orchestrating multi-agent workflows with bounded costs and structured outputs.

submit_smoke · Orchestrating multi-agent workflows with bounded costs and structured outputs.

fan-out-reduce · Orchestrating multi-agent workflows with bounded costs and structured outputs.

canary · Post-deployment health monitoring with parallel metric collection and automated judging.

mishkan-knowledge-gap-discovery · When auditing whether expected knowledge exists or needs research to fill gaps.

lets-backlog · When domain experts need to surface backlog gaps and cluster related themes in parallel.

loop-until-budget · When a task needs to run repetitively until completion or token budget exhaustion.

test · When running tests needs detailed instrumentation and result aggregation.

fan-out.template · When executing parallel agents on the same arguments without manual orchestration.

jira-digest · When generating team digests from Jira activity with structured, parseable output.

fathomdb-0.8.0-slice-plan · Planning large multi-phase database implementations with complex dependency sequencing and human-in-loop gates.

comms-workflow · Continuous monitoring and triage of PR comments with automated fixes and intelligent re-request logic.

sefer-release-notes · Assembling release notes from commits when different categories need orthogonal specialist perspectives.

playwright-parallel-test · Running parallel UI tests across form/nav/layout when isolation and concurrent feedback matter.

deep-research · Running multi-source research with web fetching, fact verification, and cited synthesis.

mine-claude-md-from-sessions · Extracting and synthesizing codebase documentation from session transcripts.

context-refresh · Refreshing and updating cached context for downstream tasks.

architecture-docs · Generating architecture documentation from code analysis and design specs.

debug · Isolate root cause of flaky tests through hypothesis-driven test mutation.

cheap-fanout-synthesis · Fan-out cheap research when no single synthesizer can hold the full pool.

claim_verifier · Verify document claims are grounded in cited sources.

api-surface-map · Discovering and documenting REST/RPC/GraphQL APIs automatically by grouping endpoints across large codebases.

reclassify-commits · Batch-labeling hundreds of commits into coherent features without manual triage.

research-digest · Aggregating research signals from 7 heterogeneous sources with real-time relevance ranking and dedup.

dag-inventory · Cataloging and classifying large numbers of DAGs for safe migration or archival decisions.

markup-repair-sonnet · Preserving localization quality while fixing structural markup mismatches token-by-token.

sync-settings-options · Keeping plugin configuration synchronized across instances without manual sync.

screenshot-analyze · Extracting semantic meaning and anomalies from screenshots via agent interpretation.

mock-schema-failure · Simulating schema failures to test error handling paths without live breaking changes.

protected-globals · Restricting variable scope in workflows to prevent unintended cross-phase mutations.