Remek
Use when a request names remek or asks to initialize a governed Agent Skills source; capture, import, or revise governed skills; record reviewed external evidence or distribution approval; audit an untrusted skill; or prepare and verify a release mirror you own. remek governs completed skill bytes and reviewed results; authoring, evaluation runs, Git, installation, and publishing use compatible capabilities.From its SKILL.md
npx -y skills add benblackthorn/remek --skill remekAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 27 days oldThe repository was created 27 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
3.4 KB, 552 tokens by cl100k_base, as published. Nobody here has run it
remek
remek governs reviewed, file-based Agent Skills represented as canonical UTF-8
file trees: provenance, policy, cases, evidence, approval, and release manifests
bound to exact bytes. Follow references/workflows.md
exactly. Run the installed scripts/cli.py with python3 -I -S -B; inside a
governed source, use its ./remek wrapper.
Start only from completed work, a design, or a reviewed import. For captured
work, write the confirmed procedure to one file the user reviews; remek
retains that exact file, never the chat. Only scaffold mutates directly.
Every remek-owned source or mirror change follows save plan → plan show →
owner approval → apply; drifted inputs refuse and nothing lands.
Never invent procedures, bypass refusals, or hand-edit governed outputs. remek never runs providers or candidates, creates repositories, installs, commits, pushes, publishes, changes visibility, or converts private history to public. remek owns governance, not the whole task. Before authoring, quietly use the host-native or known preferred compatible capability; if several credible choices remain without a preference, ask once. Give remek the completed candidate, design, or reviewed import, then continue authorized adjacent work. Never narrate capability discovery.
Bound residue per the reference. Protected artifacts use one external mode-0700
session; ignored, canonical <git-root>/.tmp/remek/<name>/<run-id>/ is
coordination-only.
Never store artifacts in installed skill directories or common personal
folders. Delete only proven disposable bytes, report retained paths and sizes,
and preserve remek boundaries for owner-selected storage.
Before first initialization, quietly discover instructions, project roots, skill repositories, Git state, and GitHub context. Reuse an established setup. Otherwise ask about existing repositories and future public skills, explain the recommended topology, and confirm the name and absolute path before creation.
Speak as the workflow and lead with the outcome. Keep routine output internal; show exact plans and refusals, then give the reference's completion status. Evidence and release approval are independent gates. The reviewer field is a declaration, not authentication or receipt attestation, and approval grants no runtime, tool, or script permission. Exposure is not installation. For release status, name audience and selected count, distinguish local verification from push readiness, report remote publication and anonymous installation as not observed by remek unless their owner separately observes them, and say when a private audience disallows anonymous installation or an empty selection releases nothing.
What ships with it: 18 files
300.1 KB alongside SKILL.md, 13 of them executable
agents/
- openai.yaml314 B
evals/
- evals.json1.3 KB
references/
- workflows.md14.5 KB
scripts/
- cli.pyruns2.0 KB
toolchain/
- assets/gate240 B
- manifest.json1.5 KB
- runtime/remek_core/app.pyruns23.3 KB
- runtime/remek_core/contract.pyruns4.5 KB
- runtime/remek_core/evaluation.pyruns11.2 KB
- runtime/remek_core/filesystem.pyruns31.3 KB
- runtime/remek_core/frontmatter.pyruns8.8 KB
- runtime/remek_core/__init__.pyruns15 B
- runtime/remek_core/model.pyruns3.4 KB
- runtime/remek_core/plans.pyruns16.2 KB
- runtime/remek_core/repository.pyruns67.8 KB
- runtime/remek_core/transaction.pyruns17.2 KB
- runtime/remek_core/workflows.pyruns92.7 KB
- scripts/cli.pyruns3.8 KB