Skip to main content

Module skills

Module skills 

Re-exports§

pub use bundle::BundleSummary;
pub use document::DocumentParseError;
pub use document::SkillDocument;
pub use frontmatter::SkillFrontmatter;
pub use reference::SkillRef;
pub use reference::SkillRefError;
pub use scaffold::ScaffoldError;
pub use scaffold::ScaffoldOptions;
pub use service::EffectiveSkill;
pub use service::EffectiveSkillSet;
pub use service::RemoveMode;
pub use service::ServiceError;
pub use service::SkillOrigin;
pub use service::SkillSummary;
pub use service::SkillsService;

Modules§

audit
bundle
Runtime-side bundle facade. The directory rules (default path, inside- shared/ containment, uniqueness) live in zeroclaw_config::skill_bundles so Config::validate and the SkillsService share one implementation. This module is a thin re-exporter plus the BundleSummary shape
cache
Process-global cache for skill-directory loads.
constants
Canonical filenames + scaffold subdirs for the Agent Skills spec.
creator
document
Parse and serialize canonical SKILL.md files.
frontmatter
Canonical SKILL.md frontmatter.
improver
reference
review
scaffold
Scaffold a new skill on disk: write SKILL.md + create optional scripts/, references/, assets/ subdirs per the canonical layout.
service
Public service surface every consumer (CLI, gateway, future TUI) uses to read and mutate skills + skill bundles. There is no second implementation — drift is closed by construction.
skill_http
skill_tool
testing

Structs§

DroppedSkill
ShadowedSkill
One lower-precedence skill that lost its name to an earlier (higher-priority) source during the agent’s effective-skill dedup. Recorded for the dashboard so operators can see why an assigned bundle skill is being overridden.
Skill
A skill is a user-defined or community-built capability. Skills live in ~/.zeroclaw/workspace/skills/<name>/SKILL.md and can include tool definitions, prompts, and automation scripts.
SkillSlashChoice
A predefined choice for a typed slash option.
SkillSlashOption
A typed option a slash-tagged skill exposes on its slash command. Shaped after the Discord Application Command Option model but channel-agnostic; a slash-capable channel maps kind to its wire option type. Declared in SKILL.toml under [[skill.slash_options]].
SkillTool
A tool defined by a skill (shell command, HTTP call, etc.)
SlashOptionKindDescriptor
Serialized capability row for one SlashOptionKind, as published to surfaces (the web dashboard mirrors this shape). Built by walking SlashOptionKind::ALL; never hand-authored.

Enums§

BundleError
SkillDropReason
Why the audited resolver dropped a candidate skill directory/file. Carries the human-readable detail the loader already logs, so the dashboard can show the same reason without re-running the audit.
SkillTier
SlashOptionKind

Functions§

build_install_tier_banner
init_skills_dir
Initialize the skills directory with a README
install_extra_registry_skill_source
Install a skill from a user-configured extra registry, addressed as registry:<name>/<skill>. The named registry must be present, enabled, and of kind = "git"; it reuses the same git-clone registry mechanism as the default bare-name registry and then installs the skill locally.
install_git_catalog_skill_source
Install a single skill by name from a git catalog repository.
install_git_skill_source
install_local_skill_source
install_registry_skill_source
is_extra_registry_source
True when source is an extra-registry spec registry:<name>/<skill> with both segments being bare registry-safe identifiers.
is_git_source
is_registry_source
load_skills
Load all skills from the workspace skills directory
load_skills_for_agent
load_skills_for_agent_audited
load_skills_for_agent plus the audit-dropped and shadowed candidates the resolver skipped, so the dashboard can surface them without re-auditing or re-walking
load_skills_for_agent_from_config
load_skills_for_agent_from_config_audited
load_skills_for_agent_from_config plus the audit-dropped and shadowed candidates the resolver skipped — the dashboard’s source for the skipped-audit banner and shadow badges
load_skills_from_directory
load_skills_with_config
Load skills using runtime config values (preferred at runtime).
load_skills_with_config_audited
Like load_skills_with_config but also returns the audit-dropped candidates the resolver skipped, so the dashboard can surface them
load_skills_with_open_skills_settings
Load skills using explicit open-skills settings.
lookup_registry_skill_tier
Look up a skill in <registry_dir>/registry.json and return its trust tier and version. Returns (SkillTier::Unknown, None) if the index file is missing, malformed, or does not list the skill.
parse_extra_registry_source
Parse registry:<name>/<skill> into (registry_name, skill_name). Returns None unless it is exactly one registry name and one skill name, both matching their install-spec identifiers.
print_install_tier_banner
Print the install-time tier banner to stdout.
skills_dir
Get the skills directory path
skills_to_prompt
Build the “Available Skills” system prompt section with full skill instructions.
skills_to_prompt_with_mode
Build the “Available Skills” system prompt section with configurable verbosity.
skills_to_tools
skills_to_tools_with_context
skills_to_tools_with_context_and_runtime
slash_option_kinds
The full registry, produced by exhaustively walking SlashOptionKind::ALL.