ioncache-ai-tools

Invoke before responding whenever the user asks a direct question. Ensures the question gets answered fully and explicitly, with reasoning, before any other output.

Checks behavioral claims against complete, current source evidence. Use before writing code comments, documentation, commit or pull request text, or answers describing code behavior, and when a claim-evidence checkpoint blocks a tool.

Use when writing or reviewing functions, for guidance on parameter counts, nesting depth, function length, and single-responsibility structure.

Use before adding a code comment, for the rule on when comments are warranted and how to write them (why, not what).

Creates a git worktree and applies the repo's .worktree-setup.json. Use when asked to create a worktree, including local config, generated caches, and post-create commands.

Use when writing or updating documentation, READMEs, commit messages, or code comments, for sentence style, punctuation, and emphasis rules.

Traces how a feature, workflow, or system works without changing files. Use when asked to investigate or explain existing code before making changes.

Use when writing or reviewing JSDoc on JavaScript/TypeScript functions, for required tags, typedef rules, and formatting. Opinionated JS/TS documentation standard.

Use whenever the task is to produce a prompt file (a .prompt.md file, or any "write me a prompt" request), for the required output format.

Reviews code for correctness, security, performance, and project standards using independent review passes. Use when asked to review files, a branch, a pull request, or a repository; reports findings without applying fixes.

Autonomously reviews, validates, fixes, and re-reviews code within finite nested loops, preserving decisions in a per-run ledger. Use when explicitly asked to run a bounded review-and-fix loop or resume one, not for a read-only review.

Use when writing or reviewing API endpoints, database access, auth, or user input handling, for validation, sanitization, and secret-handling standards. Examples assume Fastify/MongoDB but the principles are general.

Groups and fixes errors by root cause after a large change. Use when many test failures or lint errors need systematic triage.

Use when writing or reviewing JavaScript/TypeScript unit tests with Vitest, for BDD structure, AAAR comments, and coverage/mocking conventions. Opinionated, assumes Vitest.

Analyzes unresolved review comments and PR-level feedback on the active PR without making changes. Use when asked to triage review feedback into severity, validity, and required actions; do not edit files, resolve threads, post replies, or run tests.