Man Hua
Community@manhua-man
Man Hua builds session-distillation and adversarial review skills that convert Cursor and Claude Code conversations into verified, reusable knowledge-base entries.
Agent Skills by Man Hua
Showing 7 vetted skills indexed across 1 GitHub repositories.
manhua-mansession-distill
Distills Cursor Composer session transcripts into verified repository knowledge base entries.
manhua-manclaude-session-distill
Distill Claude Code JSONL session transcripts into packets, notes, and verified knowledge.
manhua-mangrill-me-docs
Audits markdown documentation against codebase entities, DTOs, and controllers for consistency.
manhua-manpacket-memory-export
Convert session-distill packets into structured draft memory entries with review queues.
manhua-manask-me
Provides architecture and roadmap consultation notes with tradeoffs, risks, and dependencies.
manhua-mananswer-me
Gathers code, doc, config, and test evidence for under-evidenced memory draft candidates.
manhua-mangrill-me
Stress-tests high-risk memory drafts and conclusions before durable knowledge promotion.
Frequently Asked Questions About Man Hua
FAQPage SchemaWhat tasks can I accomplish with Man Hua's skills?▼
You can distill Cursor Composer and Claude Code sessions into packets, session notes, and knowledge-base entries; pressure-test markdown documentation against real code entities like DTOs and controllers; gather evidence for weak memory drafts; and export structured memory candidates for Codex-mem synchronization.
Who are these skills designed for?▼
They target developers using Cursor Composer or Claude Code who want to convert AI pair-programming conversations into durable, verified project knowledge. They suit engineers maintaining documentation accuracy and teams enforcing evidence-based review before promoting conclusions into a knowledge base.
How does the session distillation and review pipeline work?▼
Run session-distill with an index, 'run --next N', or 'deep-distill --project <p>' to produce packets. The answer-me skill then verifies evidence gaps, grill-me applies adversarial scrutiny to high-risk drafts, and packet-memory-export generates structured memory entries. Unverified batch promotion is explicitly forbidden.
What permissions and dependencies do these skills require?▼
The skills operate through declared allowed-tools: Bash, Read, Write, Edit, Glob, and Grep. Distillation and export skills use the full set, while ask-me, answer-me, grill-me, and grill-me-docs are read-only collaborators restricted to Read, Grep, and Glob for consultation and auditing.
Do these skills replace human review of knowledge promotion?▼
No. ask-me and answer-me are optional collaborators for design questions and evidence gathering only; promotion decisions stay in the review chain. grill-me exists specifically to pressure-test high-risk drafts before they become durable knowledge, enforcing a verification-first policy.