Expand description
Skill manifest — full serde representation of canonical skill.yaml.
Structs§
- Content
- Governance
Ref - Governance identification for Commander integration.
Current code: serde-only seam, never read at runtime.
Commander reads
org_id+constitution_hashto load the applicable constitution and derive policy. Never stores policy here — policy belongs in the constitution, not the manifest. - Procedure
- Procedure
Step - Requirement
- Retry
Config - Commander extension: retry configuration for a workflow step.
- Skill
- Top-level skill — wraps the manifest with security metadata that lives alongside (but separate from) the publisher-authored fields.
- Skill
Manifest - Publisher-authored fields. This is what gets signed and is the unit of content hashing.
- Trigger
- Variable
Enums§
- Failure
Action - What to do when a workflow step fails.
- Skill
Scope - Visibility scope for a skill — determines which layers can see and use it.
- VarType
- Variable types for workflow/skill parameters (v2 resolved decision #3:
ONE
Variabletype lives here;workflow::Variablere-exports it). - Visibility
- Progressive disclosure: whether a skill appears in the always-injected learning index or is loadable on demand only.
Functions§
- scope_
visible - Is a skill with this (scope, fleet, project, team) visible in the given active context?
Layers combine: user/enterprise are always visible; fleet/project/team are visible
only when their selector matches the active context. (specific wins; see spec §6)
Wired into injection via
mur-coreretrieve::skill_candidates::filter_by_scope: project context is auto-derived from the cwd repo root; fleet context remains env-only until the fleet runtime supplies it; team is set fromFleet.team_id.