Indium-AI-Labs/indium-agentkit

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Audits UI markup, components, and styles against WCAG 2.2 Level AA (semantics, accessible names, keyboard and focus, contrast, target size, forms, and status messages) and maps each finding to a success criterion. Use proactively after UI changes or before releasing pages, forms, dialogs, or design-system components.…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Reviews multi-agent designs, orchestration code in any framework, and agent configuration (subagent definitions, AGENTS.md, CLAUDE.md) for unbounded loops, missing termination, over-broad tool permissions, context bloat, and prompt-injection boundaries. Use when agents, subagents, tool permissions, or orchestration…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Designs or reviews one bounded API contract (resources, operations, auth, typed requests and responses, errors, pagination, idempotency, versioning) from requirements and existing conventions, and returns an api-contract handoff. Use when a feature needs a new or changed API before implementation, or when a proposed…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Implements one bounded server-side slice (endpoint, handler, service, validation boundary, or backend test) against an agreed API contract, then returns verifiable evidence. Use when the contract and file scope are already decided. Edits only assigned paths.

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Diagnoses a CI run by tracing the run identity, job graph, and failing step, reproducing the exact command locally, and ranking causes with evidence. Use when a CI check fails, flakes, or is unexpectedly skipped, or a required check behaves wrongly. Not for full infrastructure posture (use the infrastructure-reviewer…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Reviews code, configuration, and data flows for compliance readiness against GDPR, SOC 2, HIPAA, or PCI DSS: regulated-data inventory, redaction in logs and telemetry, encryption and key management, access control and tenant isolation, audit logging, and retention and erasure. Use before an audit or a launch that…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Reviews data pipelines and warehouse models (batch, streaming, ELT, dbt-style transformations) for grain, idempotent replay, incremental and late-data correctness, quality assertions, PII handling, and cost. Use when a pipeline or published model is added or changed, or when data is wrong, late, or duplicated.…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Designs or reviews data models, schemas, keys, constraints, indexes, and access paths for a requested change, and returns the design part of a data-migration plan. Use when a feature adds or changes tables, relationships, constraints, or indexes, or when a schema or ORM model needs review before implementation. Not…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Audits dependencies from manifests and lockfiles for known vulnerabilities, reachability, license risk, staleness, and supply-chain signals without installing anything, and returns a dependency-audit report. Use before a release, after lockfile changes, or when an advisory needs impact evidence. Not for performing…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Writes or updates project documentation, including READMEs, tutorials, how-to guides, reference pages, API references, and architecture decision records, from code evidence, running or marking every command it documents. Use when documentation is missing or stale, or must ship with a change. Edits only assigned paths.

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Reviews eBPF programs (XDP, tc, kprobe, tracepoint, fentry, LSM), their maps, and their loaders for verifier rejection risk, packet and pointer safety, kernel-version portability, and dangerous helper use, by static analysis only. Use when eBPF C or Rust sources, map definitions, or loader code change, or a program…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Sizes a proposed change by splitting it into slices, giving three-point PERT estimates as ranges with stated confidence, and anchoring them in similar past changes from the repository history. Use before committing to scope or a timeline, or to decide whether a spike should come first. Read-only.

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Maps repository structure, entry points, dependencies, and data flows for a question or area. Use proactively before planning changes in unfamiliar code. Read-only.

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Reviews formal specifications and proofs (TLA+ with TLC or Apalache, Alloy, Dafny, SMT-LIB, Lean, Coq) and runs checkers within limits the parent sets, reporting counterexamples and exactly what was proven versus only model-checked. Use when a protocol, concurrent algorithm, or invariant needs checking against a spec…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Implements one bounded user-interface slice (component, page, form, or client data flow) in the project's existing framework and design system, with tests. Use when the UI behavior, API contract, and file scope are agreed. Edits only assigned paths.

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Coordinates incident response: establishes impact and severity, keeps a UTC timeline that separates facts from hypotheses, assigns investigation, drafts communications, and proposes reversible mitigations with stop plans for the parent to authorize. Use when an active or suspected production incident needs…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Audits embedded firmware and RTOS code (FreeRTOS, Zephyr, ThreadX, bare metal) for interrupt-priority and ISR safety, task stacks and memory budgets, watchdog design, DMA and cache coherence, driver robustness, and OTA signing, anti-rollback, and fallback, by static analysis only. Use when firmware, RTOS…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Reviews LLM evaluation designs, datasets, graders, and results for whether they support the decision: success criteria, judge calibration, variance and confidence intervals, contamination, personal data sent to third-party APIs, and cost. Use when an eval suite, benchmark result, or model or prompt comparison needs an…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Reviews whether an open-weight model can run locally as planned: weight and KV-cache memory math, usable GPU or unified memory, quantization tradeoffs, engine choice among Ollama, llama.cpp, and vLLM, server configuration, and streaming time to first token. Use before downloading weights or starting a local server, or…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Plans rollout, backward compatibility, and rollback for schema, data, API, and configuration migrations, and audits migration files for lock hazards, deploy-ordering breaks, and unsafe rollbacks. Use before merging or running a migration, or when a rename, type change, or breaking change must ship while old versions…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Audits iOS, Android, React Native, Expo, and Flutter release configuration (manifests, privacy manifests, permissions, signing, versioning, shrinking and symbols, OTA channels, and staged rollout readiness) for store rejection and release risk. Use before a store submission or when native build or release settings…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Measures one performance target against a reproducible baseline with approved local profiling on non-production targets, then reports bottlenecks with evidence and the next experiment. Use when a latency, throughput, CPU, memory, or startup regression needs evidence, or an optimization needs before-and-after…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Audits release readiness for a commit range: SemVer bump correctness, changelog completeness against Keep a Changelog, migration steps for breaking changes, manifest and artifact versions, and rollback readiness. Use before tagging, publishing, or approving a release candidate. Not for drafting release notes (use the…

Indium-AI-Labs/indium-agentkit

Agent Claude Code

Reviews failure handling across service, database, queue, cache, and provider boundaries: timeouts, retries and backoff, idempotency, circuit breakers, pools, queues, rate limits, fallbacks, graceful shutdown, and recovery paths. Use proactively before merging changes to dependency clients or failure handling, or when…