Feature
Advisory review
An optional, org opt-in LLM review of package content that reports findings and never gates publishing.
When an organization turns it on, the registry sends a package version’s content to the configured model (Anthropic by default, or OpenAI) and stores advisory findings with severity, category, line range and description.
Findings are advice. Only the scan result and an admin decision affect the moderation queue.
More features
All features →- 01
Validate the manifest and its trust gates
`wl validate` checks workloom.workspace.json against the schema, then runs the trust gates over a render plan and lists what they would block.
- 02
Deterministic render plans
See every file each provider will get, with its projection policy, ownership and write mode, before anything is written.
- 03
Sync with ownership checks
`wl sync` writes only the projections the trust gate allows, atomically, and changes or removes only files Workloom owns.