Problem Statement
Our catalog documentation is inconsistent about what users need to vet. docs/reference/extensions.md says to vet install-enabled catalogs as well as extensions, but preset guidance demonstrates install_allowed without equally clear catalog-vetting advice. Other install and run guides do not consistently point out that an unfamiliar project can supply its own catalog configuration.
There is also an incorrect claim in workflows/PUBLISHING.md: it describes a security review of submitted workflows and calls them “verified.” Maintainers check community catalog entries for form and completeness only. They do not review or audit the workflow code—or the code of other community components.
Proposed Solution
- Put concise guidance near the relevant catalog and install/run instructions for extensions, presets, workflows, and bundles: inspect the active catalog in an unfamiliar project, and vet both the source and the item before using it. For bundles, this includes component catalogs and their components.
- Clarify that a project-supplied catalog URL or install permission is not evidence of a code audit. Keep the distinction between first-party or deliberately curated sources and open community listings.
- Correct
workflows/PUBLISHING.md so its description of community submission checks covers entry form and completeness only, with no claim of a code or security review. Align any other conflicting review language.
Alternatives Considered
Repeating a warning at every mention of a catalog would obscure the guidance. Place it where users choose a catalog, install an item, or run content.
Component
Documentation
AI Agent (if applicable)
All agents
Use Cases
A user working in an unfamiliar project needs to recognize that project configuration may choose an install source. A user browsing a community listing needs to understand that its presence does not mean maintainers reviewed the component’s code.
Acceptance Criteria
Additional Context
The extension trust-model section and community listing notices already provide much of the intended guidance. This issue is about making that guidance consistent, not changing installation behavior or claiming that opening a project executes catalog content.
AI Disclosure
Drafted by GitHub Copilot (model: GPT-6 Sol; reasoning effort not specified) at @mnriem's direction. The issue wording and documentation assessment were AI-drafted.
Problem Statement
Our catalog documentation is inconsistent about what users need to vet.
docs/reference/extensions.mdsays to vet install-enabled catalogs as well as extensions, but preset guidance demonstratesinstall_allowedwithout equally clear catalog-vetting advice. Other install and run guides do not consistently point out that an unfamiliar project can supply its own catalog configuration.There is also an incorrect claim in
workflows/PUBLISHING.md: it describes a security review of submitted workflows and calls them “verified.” Maintainers check community catalog entries for form and completeness only. They do not review or audit the workflow code—or the code of other community components.Proposed Solution
workflows/PUBLISHING.mdso its description of community submission checks covers entry form and completeness only, with no claim of a code or security review. Align any other conflicting review language.Alternatives Considered
Repeating a warning at every mention of a catalog would obscure the guidance. Place it where users choose a catalog, install an item, or run content.
Component
Documentation
AI Agent (if applicable)
All agents
Use Cases
A user working in an unfamiliar project needs to recognize that project configuration may choose an install source. A user browsing a community listing needs to understand that its presence does not mean maintainers reviewed the component’s code.
Acceptance Criteria
Additional Context
The extension trust-model section and community listing notices already provide much of the intended guidance. This issue is about making that guidance consistent, not changing installation behavior or claiming that opening a project executes catalog content.
AI Disclosure
Drafted by GitHub Copilot (model: GPT-6 Sol; reasoning effort not specified) at @mnriem's direction. The issue wording and documentation assessment were AI-drafted.