Indium-AI-Labs

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Restructure code (extract, inline, rename, move, split, simplify, decouple) to improve clarity and maintainability while preserving observable behavior, proven by tests run after every small step. Use when code works but is hard to change, read, or test. Not for changing behavior or fixing bugs (use the…

Indium-AI-Labs/indium-agentkit

Skill Codex

Write accurate, user-facing release notes and CHANGELOG entries from the merged pull requests and commits in a release range, using Keep a Changelog categories and a breaking-changes callout with upgrade steps, with every claim checked against the diff. Use when cutting a release, drafting a GitHub release, or…

Indium-AI-Labs/indium-agentkit

Skill Codex

Resolve Git merge, rebase, or cherry-pick conflicts by recovering what each side intended, combining compatible changes, regenerating lockfiles and generated files properly, and proving the result builds and passes tests. Use when a merge, rebase, or cherry-pick stops with conflicts. Not for reviewing the merged…

Indium-AI-Labs/indium-agentkit

Skill Codex

Review a local diff, branch, commit range, or pull request for correctness, regressions, security, error handling, performance, and test coverage, verifying every finding before reporting it. Use when asked to review code or before merging a change. Not for fixing the issues found (do that only when asked) or a full…

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Implement or change FreeRTOS or Zephyr firmware: tasks, ISRs, peripheral drivers (SPI, I2C, UART, CAN), DMA, watchdogs, and memory budgets, verified with cross-compiled builds, size checks, and static analysis. Use when writing or modifying embedded RTOS code for Cortex-M, RISC-V, or ESP32 targets. Not for test rigs…

Indium-AI-Labs/indium-agentkit

Skill Codex

Plan and run a staged migration of a database schema, API contract, configuration, storage engine, or file format using expand-migrate-contract, with reconciliation evidence per stage and approval-gated rollback. Use when changing something that existing data, deployed versions, or consumers depend on, such as…

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Respond to a leaked credential (API key, token, password, or private key) committed to git, pasted in logs, or shipped in a build artifact: revoke or rotate it first, size the blast radius from audit logs, remove it from the current tree, rewrite history only with approval, and add prevention. Use when a secret was…

Indium-AI-Labs/indium-agentkit

Skill Codex

Review a scoped code change, endpoint, integration, configuration, or LLM or agent feature for exploitable security risks by tracing untrusted input to sinks and identity to protected actions, then report an evidenced findings register. Use when a change touches authentication, authorization, input handling, secrets…

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Write deterministic seed scripts that fill development, test, demo, or load-test databases with synthetic, referentially intact data that contains no real personal data, with realistic skew and checked integrity. Use when a project needs seed data, fixtures, demo data, or data volume for performance tests. Not for…

Indium-AI-Labs/indium-agentkit

Skill Codex

Investigate a bug, failing test, regression, unexpected output, or production issue by reproducing it, tracing the divergence, testing competing hypotheses, and locking the root cause with a regression test before fixing. Use when behavior is wrong and the cause is not yet proven. Not for planned refactors (use the…

Indium-AI-Labs/indium-agentkit

Skill Codex

Implement a behavior change or bug fix test-first: write a failing test at a public seam, confirm it fails for the right reason, make the smallest change that passes, then refactor with the suite green. Use when the desired behavior can be stated as an executable test before coding. Not for exploratory spikes (use the…

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Integrate an external HTTP API or webhook provider safely with a thin client that owns auth, timeouts, bounded retries with jittered backoff, idempotency keys, rate limits, and error mapping, plus webhook signature verification, replay protection, and deduplication. Use when adding or hardening a payment, messaging…

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Build a threat model for a planned feature, service, or architecture change by mapping assets, data flows, and trust boundaries, enumerating STRIDE threats per boundary, scoring likelihood times impact, and assigning testable mitigations and residual-risk owners. Use before implementing or redesigning anything that…

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Upgrade a framework, runtime, or library across a major version (for example React 18 to 19, a Node.js LTS bump, a Django or Spring Boot major, a new Python minor) by reading every intermediate migration guide, upgrading in small verified increments, applying vendor codemods, and recording behavior changes. Use when…

Indium-AI-Labs/indium-agentkit

Skill Codex

Verify finished changes with the project's own test, lint, type-check, and build commands, audit the diff for secrets and stray files, then commit, push, or open a pull request according to the repository's rules and the user's authorization. Use when work is done and needs to be checked and delivered. Not for…

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Write or update project documentation from code evidence: READMEs, tutorials, how-to guides, reference pages, explanations, and architecture decision records, running each documented command or marking it unverified. Use when asked to document a feature, API, CLI, configuration, or design decision, fix stale or broken…

Indium-AI-Labs/indium-agentkit

Skill Claude CodeCodex

Write or update an operational runbook for one failure mode of a service, with detection, diagnosis, mitigation, recovery, and escalation steps extracted from code, configuration, alerts, and drill evidence. Use when a service, alert, or failure mode has no runbook, an existing runbook is stale or untested, or a…