references
Skill26 starsChanged 4 months agoApache-2.0View on GitHub
Without it, reads hit the published release; with it, the draft (current, possibly unpublished) version. They differ whenever a skill has been edited but not republished, so pick deliberately. An unpublished skill answers 404 skill not found without --draft. Permissions differ too: draft reads want view/modify, published reads want execute/public_access/view. content/read-file return JSON holding a pre-signed object-store URL whose host (minio.…svc.cluster.local) only resolves inside the cluster — useless from a laptop. --raw writes the file's own bytes to stdout instead,…
8 other projects use a version of this.
Adopted by
Most recently taken up by
- jmche/skills-hubtheir own version, first seen Oct 20, 2026
- arthur-plautz/research-toolkittheir own version, first seen Oct 15, 2026
- darellchua2/civiltekk-opencode-claude-skillstheir own version, first seen Sep 18, 2026
- gohypergiant/agent-skillstheir own version, first seen Sep 9, 2026
- alfredxw/denovatheir own version, first seen Jul 15, 2026
- howells/fictiontheir own version, first seen May 22, 2026
Other versions8 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 8
- alfredxw/denova★ 990same name
- gohypergiant/agent-skills★ 25same name
- howells/fiction★ 23same name
- jeffsenso/dolibarr-skills★ 9same name
- v-open-platform/ai-app-skills★ 8same name
- darellchua2/civiltekk-opencode-claude-skills★ 7same name
- arthur-plautz/research-toolkit★ 0same name
- jmche/skills-hub★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository4 other files it gives its agents.
- AGENTS.mdAGENTS.md
- create-bknSkill
- create-skillSkill
- openbknSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 54
Check mocked in the response: true means the sandbox stubbed the run and the skill never executed.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
skill — registry / market / lifecycle
| Command | Notes |
list / market / get <id> / market-get <id> | Browse registry + market. Filters: --name --category --create-user --sort-by create_time|update_time|name --sort-order asc|desc --limit n（1–100，默认 30） --page n --all; list also takes --status unpublish|published|offline|editing (the market is published-only, so it has none). |
names <id...> | Resolve ids → names in one call; unknown ids are skipped, not an error. |
content <id> [--raw] [--draft] / read-file <id> <rel-path> [--raw] [--draft] / history <id> | Progressive read + versions. |
files <id> [path] [--tree] [--draft] | File listing. No path = root; a directory path = its direct children; --tree = whole hierarchy under path (or the whole skill when path is omitted). |
execute <id> --entry '<shell>' [--timeout <s>] [--raw] [--exit-code] | Run the skill in the platform sandbox. Omit --timeout to leave the limit to the sandbox (300s by default, 3600s max); the client waits out the sandbox's maximum either way. |
set-status <id> <status> | published | offline only — the endpoint refuses unpublish / editing, which come from registering and editing. |
register <dir> [--source custom|internal] [--category <c>] [--extend-info <json>] | Zip a local skill dir → multipart register. SKILL.md must have frontmatter (name/description). --source defaults to custom, matching the backend's own default (default:"custom" validate:"oneof=custom internal"), so the registered result is unchanged from omitting it. |
download <id> [out.zip] [--draft] / install <id> [dir] | Save archive / download + unzip. |
update-metadata <id> --body <json> / update-package <id> <dir> | Replace metadata / replace package (zip). Metadata is a full overwrite: name, description, category are required; source is custom|internal; extend_info is an object. The CLI refuses anything else before sending. |
republish <id> --version <v> / publish-history <id> --version <v> | republish copies a historical version back into the draft and publishes nothing; publish-history publishes that version directly. |
--draft — which version you read
Without it, reads hit the published release; with it, the draft (current,
possibly unpublished) version. They differ whenever a skill has been edited but
not republished, so pick deliberately. An unpublished skill answers 404 skill
not found without --draft. Permissions differ too: draft reads want
view/modify, published reads want execute/public_access/view.
--raw — file text instead of a link
content/read-file return JSON holding a pre-signed object-store URL whose
host (minio.…svc.cluster.local) only resolves inside the cluster — useless
from a laptop. --raw writes the file's own bytes to stdout instead, so
read-file <id> <path> --raw > file reproduces the file byte for byte. Binary
files are refused with a pointer to install, never dumped as mojibake.
--raw and --draft compose. The CLI first asks for the body inline. Draft reads
(--draft, the management routes) answer with the body inline; published reads
answer with only a URL (verified on 14.103.77.23), so the CLI fetches the archive
once per run and serves every later read from it.
Running a skill
openbkn skill execute <id> --entry 'python scripts/generate_page.py --list-styles'
openbkn skill execute <id> --entry 'python scripts/generate_page.py --style brutalist' --raw > page.html
openbkn skill execute <id> --entry 'python check.py' --exit-code # exit status = sandbox'sThe platform uploads the package into a sandbox session and runs --entry from
its work dir, so scripts reach bundled files by relative path
(styles/brutalist/tokens.json). --raw writes the run's stdout straight
through (stderr stays on stderr) so output can be redirected into a file;
--exit-code is opt-in, since it changes what the process returns. A sandbox
that reports mocked=true exits 125: nothing failed, but nothing ran either,
and --exit-code && next-step must not read that as a pass. An exit code the
shell cannot represent (>255) becomes 1 rather than truncating to 0.
Check mocked in the response: true means the sandbox stubbed the run and the
skill never executed. --raw prints that warning to stderr so it can't
contaminate a redirected capture.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.