Skills

Agent skills repository for use with any agentic tool that supports skill manifests — Devin CLI, Claude Code, Cursor, OpenCode, Codex, Gemini CLI, and others.

Each subdirectory under skills/ is a self-contained skill with a SKILL.md manifest describing its name, description, and instructions.

Included skills

Skill	Description
browser-automation	Browser automation — QA, element interaction, screencapture, network capture, cross-account/app/memory/history access.
caveman	Compression communication mode. Drops filler, hedging, and narration while keeping full sentences and technical accuracy.
code-work	Code reading and editing conventions — scoped reads, minimal edits, no comments.
commit-and-push	Commit and push current changes.
context-gathering	MANDATORY pre-flight before any codebase exploration, reading, editing, or answering about code.
database	Database inspection tools for schema, queries, and operations.
debug	Troubleshoot and fix a user-reported bug — grill to clarify, investigate root cause, plan, and execute the fix.
adjust	Apply user-requested adjustments to existing code — fixes, improvements, refactors per notes. No shipping.
knowledge-graph	Turn any input (code, docs, papers, images, videos) into a persistent knowledge graph with community detection and query/path/explain tools.
grill-with-codebase	Grill the user about a task or intent using the codebase as ground truth until decisions form a concrete plan.
implement	Develop from a ticket — load details, gather context, implement on master.
container-cli	Container CLI usage reference — start/stop/restart, exec, logs, run scripts, share tunnels.
draft-reply	Draft a reply to a ticket and post it as a private note for review. Read-only.
memory-usage	Read or write persistent project memory before, during, or after non-trivial work.
new-branch	Create and switch to a categorized branch summarizing current uncommitted or ahead-of-base work.
pr-create	Create a pull request for the current branch.
pr-fix	Address PR feedback by making fixes and responding directly to review threads.
pr-fix-ci	Fix failing CI checks on a PR — read logs, fix the code, push, verify green.
session-to-memory	Extract retention-worthy knowledge from a session and persist it to the active memory provider.
setup-skills	Guide users through initial configuration — creates <project-root>/docs folder and generates all spec files needed by other skills.
ship-changes	Ship current work end-to-end — branch, commit, push, open PR.
spawn	Invoke a target skill inside a subagent so its content stays isolated to the subagent's context.
subagents	Subagent spawning rules for parallel work, isolated tasks, and context offloading.
to-specs	Turn a request or ticket into a spec, then sync to the ticket system. Read-only.
triage	Triage a ticket — load context, gather evidence, post findings. Read-only.
using-code-navigation	Use code navigation tools for symbol-level code work instead of grep/read.
using-terminal	MANDATORY pre-flight before sending any command to the terminal/shell.

Tool Specifications (<project-root>/docs)

Skills are tool-agnostic. They reference tool specifications in a <project-root>/docs folder that lives in the project root of each consuming project — not in this skills repo.

Each project using these skills must create a <project-root>/docs folder with markdown files documenting the interface and expected behavior of the tools available in that project. Skills reference these specs by path (e.g., <project-root>/docs/issue-trackers.md), allowing agents to accomplish workflows using any tool implementation that conforms to the documented interface.

Path resolution

<project-root> is a placeholder. Before reading or passing any <project-root>/docs/... path to a tool, substitute it with the consuming project's root — resolved via git rev-parse --show-toplevel, or fall back to the agent's current working directory when not inside a git repo. Never resolve <project-root> to the filesystem root /; the leading-slash form (/docs/...) is intentionally avoided because agents misread it as a machine-root absolute path.

Expected spec files

File	Covers
<project-root>/docs/issue-trackers.md	Issue/ticket tracking tools — load, sync, list tickets
<project-root>/docs/bug-trackers.md	Bug tracking and error monitoring tools — query errors, stack traces
<project-root>/docs/mcp-servers.md	MCP server interaction — list tools, call tools, discovery protocol
<project-root>/docs/git-hosts.md	Git hosting platforms — PRs, reviews, issues, branches, commits
<project-root>/docs/database-tools.md	Database inspection — schema, queries, REPL operations
<project-root>/docs/code-navigation.md	LSP-backed code navigation — find symbols, trace references, refactor
<project-root>/docs/browser-automation.md	Browser automation — snapshots, element interaction, downloads
<project-root>/docs/knowledge-graphs.md	Knowledge graph tools — build, query, path, explain
<project-root>/docs/container-clis.md	Container CLI wrappers — lifecycle, exec, logs, scripts
<project-root>/docs/memory-providers.md	Persistent memory — list, read, write, edit, delete memories
<project-root>/docs/terminal-wrappers.md	Terminal token compression wrappers — command prefixing, output filtering
<project-root>/docs/doc-lookup.md	Documentation lookup — framework/library/SDK docs, search fallbacks

Each spec file must define: the operations available, their input/output schemas, and expected behavior rules. Agents read these files to learn how to interact with the project's specific tool implementations.

Installation

Via npx skills (recommended)

Install all skills to your agent of choice:

npx skills add danielsuguimoto/skills

Install specific skills:

npx skills add danielsuguimoto/skills --skill caveman --skill code-work

Install globally to a specific agent:

npx skills add danielsuguimoto/skills -g -a claude-code -y

List available skills without installing:

npx skills add danielsuguimoto/skills --list

Manual

Clone the repository:

git clone git@github.com:danielsuguimoto/skills.git

Then symlink or copy individual skill folders from skills/ into your agent's skills path. Common locations:

Tool	Path
Devin CLI	~/.agents/skills/
Claude Code	~/.claude/skills/
Cursor	~/.cursor/skills/
OpenCode	~/.config/opencode/skills/
Codex	~/.codex/skills/

License

See individual skill folders for license information.