@tw93

hunt enforces diagnostic discipline by blocking fixes until root cause is identified and stated with supporting evidence, reducing patch-induced regressions.

health audits agent configuration and ai maintainability against a five-layer framework (config, instructions, tools, verifiers, maintainability), producing prioritized findings with concrete evidence and actionable diagnostics.

read fetches urls and pdfs as clean markdown, with platform-specific routing (feishu, wechat, github, twitter) and fallback to proxy cascade. outputs markdown only, no analysis.

planning skill that converts rough ideas into validated decision-complete recommendations with concrete goals, constraints, and handoff steps before execution begins.

check reviews diffs and pull requests before shipping, validates against project context and ci state, and surfaces blockers or safe fixes. incomplete procedure and vague worktree interaction details limit actionability.

design produces usable interfaces with intentional visual direction, anchored to outcome verification via rendered artifacts and viewport testing. covers layout coherence, typography, and responsive states.

write removes ai-flavor patterns from prose while preserving author intent and factual claims. targets content destined for human readers across product docs, marketing, and localization work.

kami is a document typesetting system with design presets (warm parchment, serif hierarchy, ink-blue accents) for resumes, landing pages, and slide decks. applies user brand profiles and version updates non-blocking.

health audits agent config alignment across instruction surfaces, tools, and verifiers. it flags mismatches between declared behavior and runtime state, with budget-aware escalation to deep inspection when drift is suspected.

Runs a six-phase research workflow that turns unfamiliar domains, source bundles, or collected material into publish-ready output. Use when users ask in any…