Decision Deliberation
Examine an open choice from five angles before committing.
View skill ↗Open work
Skills and working examples.
Small, reusable skills for more reliable agent work, alongside two public labs that show how evidence and permission change what an agent can do.
Public repository
Agent Skills
Public skills in an open Markdown format. Each includes instructions, an overview, and a worked example.
Use one skill when the work needs it.
The collection covers decisions, evidence, handoffs, evaluation, and writing for retrieval. It is designed for agents that support SKILL.md, and the instructions can also be pasted into a chat.
Install the collection: npx skills add Kaagemusha/agent-skills
For everyday work
Grounded Research
Investigate an open question with primary evidence, keeping facts, inference and uncertainty separate.
View skill ↗Source Brief
Turn a long source into a short brief checked against the original.
View skill ↗Brief to Work Order
Give a task bounded scope, ownership, deliverables, and acceptance checks.
View skill ↗Evidence Check
Separate supported claims from inferences, assumptions, and unknowns.
View skill ↗Red Team Review
Find the few failure modes that could change a decision.
View skill ↗Handoff
Pass on unfinished work with enough context to verify and continue it.
View skill ↗Nutshell
Explain substantial completed work in one honest, plain-language paragraph.
View skill ↗For building agent systems
Eval Viability Check
Decide whether an evaluation is worth building, needs redesign, or should stop.
View skill ↗Retrievable Writing
Write notes that still make sense when only one section is retrieved.
View skill ↗Skill or Not
Test whether a repeated prompt belongs in a skill, a saved prompt, or nowhere.
View skill ↗Local-first workbenches
Five tools for clearer agent work.
Small, focused utilities for inspecting knowledge, assembling context, comparing runs, improving workflows and checking evidence. Your files stay in your browser.
Vault Lens
Map a Markdown export, find unresolved links and compare snapshots.
Open tool ↗Context Pack
Assemble complete source sections to an exact byte budget.
Open tool ↗Run Compare
Compare configurations on shared cases without hiding required failures.
Open tool ↗Workflow Trim
Calculate critical paths, effort and cost from explicit dependencies.
Open tool ↗Source Trail
Attach exact passages to claims and retain source-version checks.
Open tool ↗Public labs
Systems you can inspect.
Two working reference implementations, with source code and interactive examples.
Context Layer Lab
See how governed records, provenance, and freshness-aware retrieval keep an agent from acting on stale operational context.
Governed Action Lab
See an agent action pass through policy checks, operator approval, and a cryptographic receipt before it can execute.