Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
| * `add`: Download a skill and install it for an AI... | ||
| * `preview`: Print the generated SKILL.md to stdout. |
There was a problem hiding this comment.
I feel that hf skills add and hf skills preview description could be harmonized now that we have more than 1 skill that can be installed
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
There are 3 total unresolved issues (including 2 from previous reviews).
Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
|
Thanks for the review @Wauplin. I've responded to each item and cut down the bloat. |
JiwaniZakir
left a comment
There was a problem hiding this comment.
The removal of --codex, --cursor, and --opencode flags from hf skills add in the docs (and presumably in the CLI implementation) is a breaking change with no deprecation path mentioned. Users relying on those flags in scripts or CI pipelines will get silent failures or unrecognized option errors after upgrading. At minimum, keeping the flags but printing a deprecation warning would be safer.
In install_marketplace_skill, the two consecutive if install_dir.exists() checks are misleading — the second block is only reachable when force=True, but that's implicit rather than explicit. Using elif force: would make the control flow clearer and prevent any future regression if the first guard is accidentally changed.
The constant GITHUB_API_TIMEOUT appears to govern requests made via get_session() to the HuggingFace Hub, not GitHub's API — the name is a misnomer that will cause confusion when someone goes looking for where GitHub calls are rate-limited or configured. Renaming it to HF_API_TIMEOUT or HUB_REQUEST_TIMEOUT would match the actual usage.
|
@JiwaniZakir thanks for the tokens, but we got this. |

This PR allows uses to add and update skills from
huggingface/skillsYou can test it like this.
The naming in
huggingface/skillsshould be simplified before we merge this. i.e.hugging-face-trackio>>trackio. That's added hereNote
Medium Risk
Adds new CLI flows that download/extract content from GitHub and overwrite local skill directories, so failures could impact user files and depend on external API behavior/network reliability.
Overview
hf skills addnow installs named marketplace skills fromhuggingface/skills(defaulting tohf-cli) instead of writing a locally generatedSKILL.md, and narrows assistant-specific installs to an optional Claude symlink.Adds
hf skills upgradeto detect and refresh installed skills when the upstream revision changes, backed by a new manifest file (.hf-skill-manifest.json) and atomic replace logic for safe overwrites.Updates the generated CLI docs and expands test coverage with a fake GitHub marketplace/session to validate install/force/upgrade behavior and edge cases.
Written by Cursor Bugbot for commit 976f2ac. This will update automatically on new commits. Configure here.