Skip to main content

Skill Tools

These 14 tools manage the skill graph — a store of reusable procedures, recipes, and how-to knowledge. Skills have steps, triggers, file patterns, and usage tracking, making them ideal for capturing repeatable workflows. Skills are mirrored to .skills/ markdown files for IDE access.

info

These tools are always available. Mutation tools (marked below) are hidden when the skill graph is set to readonly.

skills_create​

Mutation — hidden in readonly mode

Creates a new skill (reusable procedure or recipe).

Parameters​

ParameterRequiredDefaultDescription
titleYes—Skill title
descriptionYes—What this skill does
stepsNo—Array of step strings (ordered procedure)
triggersNo—Array of trigger phrases that should activate this skill
inputHintsNo—Array of expected inputs or prerequisites
filePatternsNo—Array of glob patterns for relevant files (e.g. ["src/**/*.ts"])
tagsNo—Array of tags
sourceNo"user"Source: "user", "learned"
confidenceNo1.0Confidence score (0-1)

Returns​

{ skillId } — the generated skill ID.

When to use​

Save a procedure you want to reuse. For instance, after successfully deploying a service, save the steps as a skill so it can be recalled next time.


skills_update​

Mutation — hidden in readonly mode

Partially updates a skill. Only send fields you want to change.

Parameters​

ParameterRequiredDescription
skillIdYesSkill ID to update
All skills_create fieldsNoAny field from skills_create can be updated

Returns​

{ skillId, updated }.


skills_delete​

Mutation — hidden in readonly mode

Deletes the skill, all its relations, proxy nodes, and mirror directory.

Parameters​

ParameterRequiredDescription
skillIdYesSkill ID to delete

Returns​

{ skillId, deleted }.


skills_get​

Fetches a skill with all its relations and metadata.

Parameters​

ParameterRequiredDescription
skillIdYesSkill ID

Returns​

{ id, title, description, steps, triggers, inputHints, filePatterns, source, confidence, tags, usageCount, lastUsedAt, createdAt, updatedAt, dependsOn, dependedBy, related, variants, crossLinks? } — includes resolved relations and usage statistics.


skills_list​

Lists skills with optional filters.

Parameters​

ParameterRequiredDefaultDescription
sourceNo—Filter by source ("user", "learned")
tagNo—Filter by tag
filterNo—Substring match on title
limitNo50Maximum results

Returns​

Array of { id, title, source, tags, usageCount, lastUsedAt }.


Hybrid semantic search over skills.

Parameters​

ParameterRequiredDefaultDescription
queryYes—Search query (natural language)
topKNo5Seed results for BFS
bfsDepthNo1BFS expansion hops
maxResultsNo5Maximum results
minScoreNo0.5Minimum relevance score
bfsDecayNo0.8Score decay per hop
searchModeNohybridhybrid, vector, or keyword

Returns​

Array of { id, title, description, source, confidence, usageCount, tags, score }.


skills_recall​

Broad skill recall with a lower relevance threshold than skills_search.

Parameters​

ParameterRequiredDefaultDescription
contextYes—Description of the current task or situation
topKNo5Maximum results
minScoreNo0.3Minimum relevance score (lower than skills_search)

Returns​

Array of { id, title, description, steps, triggers, source, tags, score, usageCount } — includes full steps so skills can be applied immediately.

When to use​

At the start of a complex task, cast a wide net to find potentially relevant procedures. skills_recall uses a lower minScore (0.3 vs 0.5) than skills_search, so it returns more results including less obvious matches. For instance: "What skills might help with setting up a new microservice?"


skills_bump_usage​

Mutation — hidden in readonly mode

Increments a skill's usage counter and updates its last-used timestamp.

Parameters​

ParameterRequiredDescription
skillIdYesSkill ID

Returns​

{ skillId, usageCount, lastUsedAt }.

When to use​

After successfully applying a skill's recipe, bump its usage. This helps surface frequently used skills in future searches.


Mutation — hidden in readonly mode

Creates a relation between two skills.

Parameters​

ParameterRequiredDescription
fromIdYesSource skill ID
toIdYesTarget skill ID
kindYesRelation type: depends_on, related_to, or variant_of

Returns​

{ fromId, toId, kind, created }.


Mutation — hidden in readonly mode

Links a skill to another skill (same-graph) or to a node in the docs, code, files, knowledge, or tasks graph (cross-graph). Omit targetGraph for skill-to-skill links.

Parameters​

ParameterRequiredDescription
skillIdYesSource skill ID
targetIdYesTarget skill ID (same-graph) or target node ID in external graph (cross-graph)
targetGraphNoTarget graph: "docs", "code", "files", "knowledge", "tasks". Omit for skill-to-skill links.
kindYesRelation type (free-form string)
projectIdNoTarget project ID (for cross-project links)

Returns​

{ skillId, targetId, targetGraph, kind, created }.

When to use​

Connect a skill to the code or docs it relates to. For instance, link a deployment skill to the configuration files it uses.


Mutation — hidden in readonly mode

Deletes a link from a skill to another skill (same-graph) or to a node in an external graph (cross-graph). Omit targetGraph for skill-to-skill links. Orphaned proxy nodes are cleaned up automatically.

Parameters​

ParameterRequiredDescription
skillIdYesSkill ID
targetIdYesTarget node ID
targetGraphNoTarget graph. Omit for skill-to-skill links.
projectIdNoTarget project ID

Returns​

{ skillId, targetId, deleted }.


skills_find_linked​

Reverse lookup: finds all skills that link to a specific node in another graph.

Parameters​

ParameterRequiredDescription
targetIdYesTarget node ID
targetGraphYesWhich graph the target is in
kindNoFilter by relation kind
projectIdNoTarget project ID

Returns​

Array of { skillId, title, kind, source, tags }.

When to use​

Check if any skills are associated with a piece of code or documentation before modifying it.


skills_add_attachment​

Mutation — hidden in readonly mode

Attaches a file to a skill.

Parameters​

ParameterRequiredDescription
skillIdYesSkill ID
filePathYesAbsolute path to the file on disk

Returns​

{ filename, mimeType, size, addedAt }.

note

Max 10 MB per file. Max 20 attachments per entity.


skills_remove_attachment​

Mutation — hidden in readonly mode

Removes a file attachment from a skill.

Parameters​

ParameterRequiredDescription
skillIdYesSkill ID
filenameYesFilename to remove

Returns​

{ deleted: filename }.