S
smithadifd/skills
Claude Code skills and agents that build engineering discipline into agentic work: plan, build, review, ship…
updated 2026-11-19
View on GitHub ↗Link checked 2026-11-17 · HTTP 200 · opens the vendor's own site
- Star count
- 0 ★ · smithadifd/skills · read from the GitHub search API · (checked 2026-11-19 20:00:48)
- Source
- read directly from its GitHub page · type: open-source project
- Harvest history
- first seen 2026-11-16 · last seen 2026-11-20 · seen 5× · listed by 1 sources
- Score
- score 103 · 1 sources · repo
FAQ (answers come from our records)
Who is it for?
We filed it under these tags: Agent skills. The tags come from the entry itself, not a taxonomy we imposed.
Who maintains it?
repository
smithadifd/skills;first catalogued 2026-11-16;last seen 2026-11-20。There is no official certification or quality rating here — we record sources and dates only.README excerpt (English, untranslated) · Read the full README on GitHub →
# Skills and agents for Claude Code
Excerpted from the repository README (abridged)
Start here: Where each skill fits says which skill to run at the start, middle and end of a session, and takes four everyday situations through them step by step: a new idea, a bug report, stopping for the day, picking up tomorrow.
## Why
Before coding agents, the practices that keep a codebase sound were held in place by people and by friction. The test came first because the team had agreed it would. A second person read the change because the pull request would not merge until they had. The plan came before the build because the build was going to take a week. Writing code was the slow part, so the checks fit around it.
An agent writes the code in minutes. The checks cost what they always did, but next to the agent they feel slow, so they get skipped, or the agent was never asked to do them and nobody noticed. The engineer's work shifts from writing the change to deciding what to build and verifying what came back, and there are more of those decisions, closer together, than there used to be.
Most mistakes in software are not hard decisions made badly. They are easy checks skipped because the moment to make them had passed. The test that would have failed was never written. The reviewer had watched the code being written and saw what was meant, not what was there. The push happened before anyone asked whether it should. This collection takes the engineering and project-management practices that used to live in people and process and builds them into the agent's own workflow, each one at the point where it is cheapest to skip, so that the careful path is also the path of least resistance. Good decisions should be the default, not an act of discipline.
What that looks like in practice:
In a ticket pass, the build does not start until you have seen the plan and said go, and nothing leaves your machine (no push, no pull request, no comment on the ticket) until you have seen the result and said go again.
In debug and tdd, a fix is not a fix until a test has failed for the reported cause and then passed.
The reviewer agent sees the diff with no memory of how it was written, and its verdict shows what it ran.
handoff ends a session with a note whose verified checks are commands, so pickup can rerun them tomorrow instead of trusting them.
None of this is new. Built on says where the ideas came from.
## What this is
Seventeen Claude Code skills and eight agents that take a change from first idea to handoff: planning, building, reviewing, shipping, and picking the work up again in a later session.
The difference is in how the collection is built. Every skill that ships a script ships its tests, about 2,700 across the repository, and the lint refuses a script without a tests/ directory. Every change is reviewed before merge by a reviewer with no context of the build, and code changes also by a second model family. A fix lands with mutation evidence in its pull request: the guarded line reverted, the test shown failing, then passing again with the fix restored. Each skill works from a plain install with no sibling checkout, and the CI check proves it by installing the whole set into an empty home directory.
## Why
Before coding agents, the practices that keep a codebase sound were held in place by people and by friction. The test came first because the team had agreed it would. A second person read the change because the pull request would not merge until they had. The plan came before the build because the build was going to take a week. Writing code was the slow part, so the checks fit around it.
An agent writes the code in minutes. The checks cost what they always did, but next to the agent they feel slow, so they get skipped, or the agent was never asked to do them and nobody noticed. The engineer's work shifts from writing the change to deciding what to build and verifying what came back, and there are more of those decisions, closer together, than there used to be.
Most mistakes in software are not hard decisions made badly. They are easy checks skipped because the moment to make them had passed. The test that would have failed was never written. The reviewer had watched the code being written and saw what was meant, not what was there. The push happened before anyone asked whether it should. This collection takes the engineering and project-management practices that used to live in people and process and builds them into the agent's own workflow, each one at the point where it is cheapest to skip, so that the careful path is also the path of least resistance. Good decisions should be the default, not an act of discipline.
What that looks like in practice:
In a ticket pass, the build does not start until you have seen the plan and said go, and nothing leaves your machine (no push, no pull request, no comment on the ticket) until you have seen the result and said go again.
In debug and tdd, a fix is not a fix until a test has failed for the reported cause and then passed.
The reviewer agent sees the diff with no memory of how it was written, and its verdict shows what it ran.
handoff ends a session with a note whose verified checks are commands, so pickup can rerun them tomorrow instead of trusting them.
None of this is new. Built on says where the ideas came from.
## What this is
Seventeen Claude Code skills and eight agents that take a change from first idea to handoff: planning, building, reviewing, shipping, and picking the work up again in a later session.
The difference is in how the collection is built. Every skill that ships a script ships its tests, about 2,700 across the repository, and the lint refuses a script without a tests/ directory. Every change is reviewed before merge by a reviewer with no context of the build, and code changes also by a second model family. A fix lands with mutation evidence in its pull request: the guarded line reverted, the test shown failing, then passing again with the fix restored. Each skill works from a plain install with no sibling checkout, and the CI check proves it by installing the whole set into an empty home directory.
Listed badge (optional)
Copy this line into the README if you want the listing mark. Optional, and it does not affect being listed.
[](https://anclum.com/en/s/smithadifd-skills/)
- Source
- github · https://github.com/smithadifd/skills
- Browse by field
- Agent skills
- added
- today
More entries
Filed under
Every tag opens the full list for that filter. Counts are the live catalogue figures.