Agent Skills

Skills you can read, then install

Each skill is published verbatim — the SKILL.md and every reference doc, rendered for review. When it fits your workflow, download the zip and drop it into ~/.claude/skills.

Standalone skills

Skill

java-memory-leak-audit

Thoroughly scan Java/Spring Boot code for memory leaks and database resource leaks (connections, statements, streaming queries, persistence-context bloat), tracing retention paths across the whole workflow — not just single files — analyze application/GC logs and heap dumps for leak evidence, then produce a severity-ranked findings report and implement the fixes. Use this skill whenever the user mentions memory leaks, OutOfMemoryError, heap growth, GC pressure, connection pool exhaustion, heap dumps, hprof files, GC logs, "heap keeps climbing", leak audit, or asks to review Java classes for memory/resource safety, even if they never say the word "leak" (e.g., "why does my service slow down over time?", "audit these classes for memory issues", "requests hang waiting for connections", "what's eating memory in this dump?").

5 reference docs · zip downloadOpen

SDLC Agent Toolkit

A compiler-backed, phase-gated SDLC framework for AI coding agents. These nine skills are the toolkit's compiled Claude Code output — one skill per lifecycle phase, from repository discovery through requirements, planning, implementation, testing, review, security, and documentation.

Download full suite

Skill

discover-repository-inspect

Map an unfamiliar repository into a cited orientation brief: stack, layout, build/test/lint commands, conventions. Use before planning or editing in a codebase not yet inspected this session.

zip downloadOpen

Skill

requirements-feature-analyze

Turn a feature request into numbered requirements and acceptance criteria in traceability.yaml, with every ambiguity and assumption recorded. Use before planning or estimating an implementation.

zip downloadOpen

Skill

plan-feature-implement

Produce the impact analysis and implementation plan for recorded requirements, tracing every task to an acceptance criterion, then stop for the approval gate. Use after requirements exist and before any edit.

zip downloadOpen

Skill

develop-feature

Implement a feature end to end behind a human approval gate: inspect, specify, plan, approve, implement, test, review, summarize. Use when a change needs traceable requirements and a recorded approval.

zip downloadOpen

Skill

develop-feature-execute

Implement the approved tasks and record which files each one produced. Use only after approvals.yaml carries an approved approve-implementation entry; never to start work the gate has not decided.

zip downloadOpen

Skill

test-change-validate

Cover each acceptance criterion with a test, keep only tests that pass without decreasing coverage, and record every validation verdict including not_run. Use after implementation.

zip downloadOpen

Skill

review-code-perform

Review a change set for correctness, performance and convention fit, emitting located findings with severity and confidence. Use after implementation and before requesting approval to merge.

1 reference doc · zip downloadOpen

Skill

security-change-review

Review a change for the security consequences a code review does not cover, emitting located findings with severity, confidence and an exploit path, filtered in two stages. Use after the code review.

zip downloadOpen

Skill

document-change-summarize

Close a run with a summary whose every claim resolves to a recorded id: what changed, which requirements are satisfied, which tests ran, what risk and which degradations remain. Use as the last step.

zip downloadOpen