Kaka
Community@c2kaka · Nanjing China
Seize the day!
Agent Skills by Kaka
Showing 21 vetted skills indexed across 2 GitHub repositories.
review-coding-agent-design
Review coding-agent design documents and code for complexity, consistency, and evidence-backed risks.
bootstrap-ai-collab-infra
Generates architecture diagrams, API catalogs, data models, and CLAUDE.md for a repository.
codex-local-saas-browser
Prepares and verifies the CogDB SaaS HTTPS dev server on localhost:3000 for the Codex in-app browser.
feature-intake
Reverse-engineers a Feature Intake Spec from HTML prototypes and backend API documentation.
jira-update
Apply guarded JIRA comments, workflow transitions, and tester assignment after a completed fix.
verify-frontend-delivery
Prepares frozen frontend delivery contracts and independently verifies changes against deterministic gates.
interview-prep-from-project
Extract resume highlights from a codebase and generate interview questions with reference answers.
ai-agent-framework-design-guide
Writes and reviews Chinese-first AI agent framework design documents with layered architecture guidance.
spec-to-executable-tickets
Convert product materials into evidence-backed specifications and dependency-ready executable tickets.
jira-auto-fix
Reproduce, diagnose, fix, and locally commit JIRA bug fixes through gated approval phases.
jira-report-quality-issue
Create evidence-backed JIRA issues with screenshots after guarded preflight and explicit approval.
video-insight-report
Generate evidence-linked offline HTML reports from public YouTube or Bilibili videos.
run-local-fullstack-debug
Start a local full-stack system and localize failures across service boundaries.
analyze-ai-agent-codebase
Analyze AI agent repositories by tracing architecture layers, contracts, and execution paths.
my-jira-query
Query JIRA issue details via REST API using local .env credentials.
write-weekly-report
Generates Chinese weekly report Excel files from Git commits with weighted hour allocation.
change-risk-review
Reviews git diffs for behavior, protocol, and architecture risks before commit.
propagate-api-contract-changes
Trace backend API contract changes through every affected frontend layer.
humanizer-zh
Detect and rewrite AI-generated text patterns into natural human writing.
wu-jun-writing-style
Simulate Dr. Jun Wu's writing style for Chinese column-style essays with logical reasoning and case studies.
writing-style-distill
Extract authors' writing styles from local corpora and online sources.
Frequently Asked Questions About Kaka
FAQPage SchemaWhat tasks can I accomplish using Kaka's skills?▼
You can fix and transition JIRA issues end-to-end, verify frontend deliveries against PRDs and API contracts, propagate backend schema changes across frontend layers, review git changes for risk before commit, generate Chinese weekly reports from commit history, and produce evidence-linked video insight reports.
Who are these skills designed for?▼
Frontend and full-stack engineers working in JIRA-managed teams, developers preparing Chinese-language interview materials or weekly reports, and engineers reviewing coding-agent designs or analyzing open-source agent codebases for architecture, execution loops, and extension points.
How do the JIRA skills work in practice?▼
jira-auto-fix uses a skill-local .env with read-only JIRA REST access to reproduce and fix bugs with explicit solution approval before edits. jira-update then adds approved comments, transitions the issue to testing, and assigns the tester using guarded writes that safely resume after partial failures.
What prerequisites or dependencies do these skills require?▼
JIRA skills need a configured server and .env credentials; codex-local-saas-browser requires macOS with certificate trust and OIDC setup for localhost:3000; spec-to-executable-tickets optionally uses a configured Ticket Queue Executor and real Chrome for acceptance runs.
Do any skills require explicit manual invocation?▼
Yes. verify-frontend-delivery, spec-to-executable-tickets, run-local-fullstack-debug, and codex-local-saas-browser only activate on explicit user invocation, while feature-intake, bootstrap-ai-collab-infra, and several others set disable-model-invocation to prevent automatic triggering.