Skip to content
verify mcp Beta VerifyMCP is currently in beta. If you notice any issues, get in touch and we’ll put it right.

Skillsmith

NPM · @SKILLSMITH/MCP-SERVER · SCANNED SEP 20

Share agent skills across your team, scan them for risk, and track what's actually used.

+15 this week 88 Trust /100
Trust breakdown (7 categories)

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. How we score → Why this is hard to score →

Supply Chain Security90
  • No malware found by supply-chain analysis.Pass
  • CVE check failed: a known high-severity CVE affects sharp 0.34.5, reached via @skillsmith/core > @huggingface/transformers > sharp. A fixed version is available. View diagnostics → Fail
  • No install/post-install scripts declared.Pass
  • Dependency health was assessed across the 300 of 301 dependencies we could resolve, so this covers what we could see, not the whole tree. View diagnostics → Partial
Provenance & Transparency87
  • Source repository is publicly reachable at the declared URL. View diagnostics → Pass
  • Cryptographically verified build provenance (signed, bound to smith-horn/skillsmith). View diagnostics → Pass
  • License check failed: the license (Elastic-2.0) isn't a recognized OSI-approved license. See how to fix → Fail
  • Actively maintained (last published 6 days ago).Pass
  • Publishes a security disclosure policy (SECURITY.md).Pass
Schema Quality & AI Usability68
  • AI-judged instruction clarity (excellent).Pass
  • Context-footprint check failed: tool/resource definitions use about 7275 tokens (~169/item across 43 items; 43 tools + 0 resources), over budget; trim descriptions and params. See how to fix → Fail
  • Usage-examples check failed: none of the tools include examples. See how to fix → Fail
Stability & Change Management93
  • Stability observed for 28 of 30 days with no destabilising changes; credit accrues until the full window elapses.Partial
Tool Coverage100
  • 100% of tools have a non-trivial description (not blank, and not just the tool's name).Pass
  • 100% of tool parameters carry a description.Pass
  • Structured output schemas are declared (7% of tools); any adoption earns full credit.Pass
Tool Safety100
  • No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.Pass
  • All 4 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation.Pass
  • An AI judge read all 43 captured unit(s) of tool text and found none that tries to manipulate the model reading it.Pass
Capabilities100
  • Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.Pass
Install

How do I install the Skillsmith MCP server?

Skillsmith runs locally as an npm package, launched with npx -y @skillsmith/mcp-server. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

npm · @skillsmith/mcp-server

# add to Claude Code
claude mcp add smith-horn-skillsmith -- npx -y @skillsmith/mcp-server
// .cursor/mcp.json
{
  "mcpServers": {
    "smith-horn-skillsmith": {
      "command": "npx",
      "args": [
        "-y",
        "@skillsmith/mcp-server"
      ]
    }
  }
}
// .vscode/mcp.json
{
  "servers": {
    "smith-horn-skillsmith": {
      "command": "npx",
      "args": [
        "-y",
        "@skillsmith/mcp-server"
      ]
    }
  }
}
# add to Codex CLI
codex mcp add smith-horn-skillsmith -- npx -y @skillsmith/mcp-server
// opencode.json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "smith-horn-skillsmith": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "@skillsmith/mcp-server"
      ],
      "enabled": true
    }
  }
}
# add to OpenClaw
openclaw mcp add smith-horn-skillsmith --command npx --arg -y --arg @skillsmith/mcp-server
# ~/.hermes/config.yaml
mcp_servers:
  smith-horn-skillsmith:
    command: "npx"
    args: ["-y", "@skillsmith/mcp-server"]
// ~/.netclaw/config/netclaw.json
{
  "McpServers": {
    "smith-horn-skillsmith": {
      "Transport": "stdio",
      "Command": "npx",
      "Arguments": [
        "-y",
        "@skillsmith/mcp-server"
      ]
    }
  }
}
# add to Vellum
assistant mcp add smith-horn-skillsmith -t stdio -c npx -a -y @skillsmith/mcp-server
// mcp.json
{
  "mcpServers": {
    "smith-horn-skillsmith": {
      "command": "npx",
      "args": [
        "-y",
        "@skillsmith/mcp-server"
      ]
    }
  }
}
Changelog

Every change we have recorded for this component, newest first. Security-relevant changes are always shown. ▲ marks a change for the better, ▼ a change for the worse; unmarked changes are neutral.

  • 20 Sept 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 90 to 93. That category is still filling its 30-day observation window: 27 days of observed history at the previous scan, 28 at this one. The score rises as the window fills, whether or not the server changes.

  • 18 Sept 26 +1
    • Security disclosure: unverified → pass functional
  • 17 Sept 26 0
    • Security disclosure: pass → unverified functional
  • 16 Sept 26 −3
    • Stability: pass → 0.80 functional
  • 15 Sept 26 +1
    • Stability: 0.97 → pass security
  • 14 Sept 26 +15
    • Malware scan: unverified → pass security
  • 13 Sept 26 −16
    • Malware scan: pass → unverified security
    • Stability: pass → 0.93 functional
    • Package version: 0.7.14 → 0.7.16 functional
    • Package version: 0.7.14 → 0.7.15 functional
  • 12 Sept 26 +1
    • Stability: 0.97 → pass security
Diagnostics

Diagnostic detail from the automated scan of this channel: what the scanner observed at each step, so you can see exactly where a check passed or failed. It is informational only and never changes the trust score.

Captured 20 Sept 2026 · Analysed npm/@skillsmith/mcp-server@0.7.16

Provenance Verified

A signed build attestation was found and verified, binding this exact artifact to the source repository it claims to come from.

Result Verified
Ecosystem npm
Reason Verified
Discovered via Registry attestation endpoint
Source repo smith-horn/skillsmith
Certificate issuer https://token.actions.githubusercontent.com
Certificate SAN https://github.com/smith-horn/skillsmith/.github/workflows/publish.yml@refs/heads/main
Rekor log index 2820802020
Predicate type https://slsa.dev/provenance/v1
Subject digest sha512:d1caca11a77dd8e1ad8846c5edfffd2841e965d010eabcfe45b65617efa76c4170e13ade29bf313828c05dd8b629208e6c5ac7a8da6a349948d81f2c4

Background: How many MCP packages publish verified provenance →

Vulnerabilities 2 findings
ID CVE Severity Vector Fix available
GHSA-f88m-g3jw-g9cj high yes
GHSA-rgj7-g3m4-5g8c high yes

Background: What a vulnerability scan can and cannot prove →

Dependencies 300 packages
Packages resolved 300
Deprecated 2
Stale 91
Tree resolution Partial

The dependency tree was only partially resolved, so these counts may be incomplete.

Background: SBOMs and build attestations, explained →

MCP tools · 43 exposed · ~7,275 tokens

The tools this component advertises to a client, with an estimated token cost for each. Expand a tool to see its parameters and schema. The per-tool counts are indicative and are not scored directly; the schema's total context footprint is one signal in Schema Quality & AI Usability. A tool's description is untrusted text the model reads on every call, which is what makes this list a security surface and not just an inventory: how tool poisoning works →

Tool Tokens
analytics_dashboard ~61

Enterprise analytics dashboard: recommendation accuracy, skill usage trends, team-wide aggregation. Requires Enterprise tier (advanced_analytics feature).

NameTypeReqDescription
includeRecommendationsbooleanInclude recommendation accuracy metrics (default false)
periodstringTime period (default 30d)

No output schema declared.

No examples provided.

api_key_manage ~104

Manage API keys for programmatic access. Keys are shown once on creation. Stored as SHA-256 hashes. Requires Enterprise tier (custom_integrations feature).

NameTypeReqDescription
actionstringyesAPI key operation
expiresInstringExpiration period (default: 90d)
keyIdstringKey ID (required for revoke/get)
namestringKey name (required for create)
permissionsarrayPermission scopes

No output schema declared.

No examples provided.

apply_manifest_reconcile ~493

[Skillsmith — Maintain stage] Repair a corrupted or ambiguous ~/.skillsmith/manifest.json entry through a supported path — MUTATES the manifest (with a pre-mutation backup and a durable, revertible ledger entry). Use after skill_outdated reports 'identity-mismatch' or 'local-drift', or after skill_recover_source found low/no confidence. Actions: 'mark_local' (stop tracking this entry against the registry — writes source:'unknown' + provenance:'local'); 'relink' (assert an explicit, registry-validated id/source pair — requires BOTH; never infers an identity); 'drop_entry' (hard-remove an entry whose installPath no longer resolves); 'verify' (re-check on-disk content against the registry's current content hash for one entry or, by default, every entry — writes verifiedAt only on a match); 'revert' (durable, cross-session undo of a prior reconcile on ONE entry, by ledgerEntryId or by name — survives an unrelated skill install happening in between).

NameTypeReqDescription
actionstringyes'mark_local' | 'relink' | 'drop_entry' | 'verify' | 'revert' — see tool description for each action's effect.
clientstringTarget client (ADR-139). Defaults to the canonical client (claude-code).
cwdstringAbsolute path to the calling client's actual project root, used as the ancestor-walk starting point for workspace scope resolution.
idstringrelink only, required together with 'source': registry skill id (author/name).
ledgerEntryIdstringrevert only: precise ledger entry id (mrc_...) from a prior mutating action response. Preferred over name when known.
namestringSkill name (manifest entry, not a full id). Required for mark_local/relink/drop_entry. Optional for verify (omit to batch-verify every entry) and revert (omit when passing ledgerEntryId).
reasonstringOptional human-readable reason, recorded in the revert ledger.
scopestringExplicit install scope (ADR-139). Defaults to auto-detecting an EXISTING workspace directory, else global.
sourcestringrelink only, required together with 'id': the source reference to record (e.g. github:owner/repo).

No output schema declared.

No examples provided.

apply_namespace_rename ~335

[Skillsmith — Maintain stage] Apply a rename suggestion from a prior local namespace-collision audit (`skill_inventory_audit`), or revert one already applied. MUTATES `~/.claude` (renames a skill/command/agent file) — but ONLY when `confirmed: true`. Without `confirmed`, returns a non-mutating preview ({ preview: true, before, after, applied: false, direction }) so the agent can show the change before committing. `action: 'apply'` uses the suggested name; `'custom'` uses `customName`; `'skip'` records a no-op; `'revert'` undoes a previously applied rename for the same auditId + collisionId — this is the durable, cross-session undo path (a fresh MCP server process, e.g. a new session, can still revert a rename applied by an earlier process, since the ledger persists to disk).

NameTypeReqDescription
actionstringyes'apply' uses the suggested name; 'custom' uses customName; 'skip' is a no-op; 'revert' undoes a previously applied apply/custom rename for the same auditId + collisionId.
auditIdstringyesauditId from a prior skill_inventory_audit response.
collisionIdstringyescollisionId of the RenameSuggestion to apply or revert.
confirmedbooleanWhen true, performs the file rename (or revert). When omitted/false, returns a non-mutating preview. Defaults to false.
customNamestringRequired when action === 'custom'; forbidden otherwise (including revert).

No output schema declared.

No examples provided.

apply_recommended_edit ~170

[Skillsmith — Maintain stage] Apply a recommended prose edit from a prior `skill_inventory_audit`. MUTATES `~/.claude` (rewrites a SKILL.md / CLAUDE.md snippet) — but ONLY when `confirmed: true`. Without `confirmed`, returns a non-mutating preview ({ preview: true, before, after, applied: false }). Gated on APPLY_TEMPLATE_REGISTRY: only registered when at least one apply-eligible template is enabled.

NameTypeReqDescription
auditIdstringyesauditId from a prior skill_inventory_audit response.
collisionIdstringyescollisionId of the RecommendedEdit to apply.
confirmedbooleanWhen true, performs the prose edit. When omitted/false, returns a non-mutating preview. Defaults to false.

No output schema declared.

No examples provided.

audit_export ~105

Export audit log events for a given time range. Requires Enterprise tier (audit_logging feature). Returns structured audit events in JSON format.

NameTypeReqDescription
endDatestringISO 8601 end date (defaults to now)
eventTypestringFilter by event type
limitnumberMax events to return (default 100, max 1000)
startDatestringISO 8601 start date (defaults to 24h ago)

No output schema declared.

No examples provided.

audit_query ~128

Query audit logs with filters (actor, resource, event type, result, date range). Requires Enterprise tier (audit_logging feature).

NameTypeReqDescription
actorstringFilter by actor (user ID or email)
endDatestringISO 8601 end date
eventTypestringFilter by event type
limitnumberMax results (default 100, max 1000)
resourcestringFilter by resource name
resultstringFilter by result
startDatestringISO 8601 start date

No output schema declared.

No examples provided.

compliance_report ~141

Generate compliance reports: SOC2 (markdown), CycloneDX (JSON SBOM), or raw JSON. Scoped to local skill inventory. Requires Team tier or higher (compliance_reports feature).

NameTypeReqDescription
backfillDependenciesbooleancyclonedx format only. Opt-in inline dependency extraction when skill_dependencies is empty (default: false, emits a pending-rescan placeholder instead). better-sqlite3 only.
formatstringyesReport format: soc2, cyclonedx, or json
includeUserActivitybooleanInclude user activity summary (default: true)
periodstringReporting period (default: 90d)

No output schema declared.

No examples provided.

configure_sso ~193

Configure SSO/SAML integration for your organization. Actions: set (store IdP config), test (connection test), remove (clear config), claim_domain (issue a DNS TXT verification token), verify_domain (check the TXT record). Requires Enterprise tier (sso_saml feature).

NameTypeReqDescription
actionstringyesSSO operation: set, test, remove, claim_domain, or verify_domain
domainstringDomain to claim, verify, or register (required for set/claim_domain/verify_domain)
idpEntityIdstringIdP entity ID (optional, extracted from metadata)
idpMetadataUrlstringIdP metadata URL (required for set)
memberDispositionstringHow to handle existing SSO-provisioned members on remove (default: convert_to_manual)
protocolstringSSO protocol (default: saml)

No output schema declared.

No examples provided.

get_skill ~177

[Skillsmith — Evaluate stage] Fetch full details for a specific Skillsmith-registry skill by ID. Use when the user wants details/info/description of a known skill — e.g. 'what does microsoft/playwright-cli do?', 'show me details for getsentry/commit', 'describe the vercel-labs/vercel-react-best-practices skill'. Returns name, description, trust tier, quality score, dependencies, compatibility, repository URL, install count, and an `also_installed` array of co-installed skills. Skillsmith is a registry for sharing, scanning, and tracking agent skills across any MCP-capable runtime.

NameTypeReqDescription
idstringyesThe skill ID exactly as returned by `search` (e.g., "getsentry/commit" or UUID). Do not guess the author segment — use a search result.
NameTypeReqDescription
also_installedarray
contentstring
dependenciesarray
installCommandstringyes
skillobjectyes
timingobjectyes

No examples provided.

index_local ~68

Index local skills from ~/.claude/skills/ directory. Returns count and details of indexed skills.

NameTypeReqDescription
forcebooleanForce re-indexing even if cache is valid (default: false)
skillsDirstringCustom skills directory path (defaults to ~/.claude/skills/)

No output schema declared.

No examples provided.

install_skill ~544

[Skillsmith — Install stage] Install an agent skill (SKILL.md format) from the Skillsmith registry or a GitHub repository to the local Claude Code skills directory (~/.claude/skills/) — or runtime-equivalent path when SKILLSMITH_CLIENT is set (cursor, copilot, windsurf). Use when the user asks to install/add/get a specific skill — e.g. 'install playwright-cli', 'add getsentry/commit', 'use Skillsmith to install the testing skill'. Performs Skillsmith security scan and content optimization (decomposition, subagent generation) before installation. Returns install path and Skillsmith optimization summary. Skillsmith is a registry for sharing, scanning, and tracking agent skills across any MCP-capable runtime.

NameTypeReqDescription
alsoLinkarrayAdditional clients to fan-out into (default: copy; pair with symlink for POSIX symlinks)
clientstringTarget agent (default: SKILLSMITH_CLIENT env or claude-code). Codex users pass agents.
confirmedbooleanConfirm install despite security warnings (required for experimental/unknown tier skills)
conflictActionstringAction when local modifications detected: overwrite (backup + replace), merge (three-way), or cancel
cwdstringAbsolute path to the calling client's actual project/workspace root, used to resolve project-scoped companion-agent output (e.g. Antigravity) correctly. Optional for flat, absolute-path clients (unus…
forcebooleanForce reinstall if skill already exists
scopestringInstall scope (ADR-139): "workspace" resolves against the nearest ancestor workspace marker or .git root (walked from cwd when provided), creating the client's workspace skills directory if none exis…
skillIdstringyesSkill ID (owner/repo/skill) or GitHub URL
skipOptimizebooleanSkip Skillsmith optimization (decomposition, subagent generation)
skipScanbooleanSkip security scan (not recommended)
symlinkbooleanUse relative symlinks instead of file copies for alsoLink targets (POSIX only; falls back to copy on Windows EPERM)

No output schema declared.

No examples provided.

inventory_push ~74

[Skillsmith — Sync stage] Push this machine's installed-skill inventory to your Skillsmith account so it appears on the web dashboard. Read-only monitoring — no local skill files are modified. Requires `skillsmith login`; respects the local SKILLSMITH_INVENTORY_DISABLE flag and your account's server-side consent setting.

Input schema present but exposes no named parameters.

No output schema declared.

No examples provided.

private_registry_manage ~344

Manage skills in your private registry (list, get, install, deprecate, undeprecate, namespace, submissions, approve, reject). Requires Enterprise tier (private_registry feature). Published versions require a team admin to approve them via "approve" before other members can install them (see "submissions" to check status).

NameTypeReqDescription
actionstringyesRegistry operation to perform. "namespace" returns your team's registry namespace (the required skill_id prefix) without attempting a publish. "install" downloads the skill and writes it to your skil…
forcebooleanReinstall over an existing install
includeDeprecatedbooleanAction "list" only: include deprecated versions (default hides them). No effect on "get"/"install", which always exclude deprecated versions, even by exact version.
notestringOptional reviewer note for action "approve"/"reject" (max 1000 chars)
skillIdstringSkill ID in author/name format (get/deprecate/undeprecate/install/approve/reject)
statusstringFilter for action "submissions"; omit for everything visible to you
versionstringVersion filter; required for approve/reject; "install" defaults to the most recently published

No output schema declared.

No examples provided.

private_registry_publish ~139

Publish a skill to your organization's private registry. Requires Enterprise tier (private_registry feature). Skills are scoped to your team's registry namespace and published versions are immutable. A published version is not installable by teammates until a team admin approves it — see private_registry_manage action "submissions" to check its status.

NameTypeReqDescription
contentobjectyesPackaged skill files as a { path: text } map; must include "SKILL.md" (max 2 MB total)
descriptionstringOptional skill description
skillIdstringyesSkill ID in author/name format
versionstringyesSemver version to publish

No output schema declared.

No examples provided.

publish_private ~80

Mark a skill as private on this machine: hides it from local community-search results. This is a local-only setting today -- it does not sync or share the skill with your teammates. For a real shared team registry, see the Enterprise-tier private_registry_publish tool. Requires Team tier license.

NameTypeReqDescription
skillIdstringyesSkill ID in author/name format

No output schema declared.

No examples provided.

rbac_assign_role ~114

Assign or revoke a team member's role, or list current members and their roles. Roles are fixed: admin or member (owner is managed separately and can never be changed here). Requires the team:manage_rbac permission and Enterprise tier (rbac feature).

NameTypeReqDescription
actionstringyesAssignment operation
memberIdstringteam_members.id from list_assignments (required for assign/revoke)
rolestringRole to assign, or the role to revoke (required for assign/revoke)

No output schema declared.

No examples provided.

rbac_create_policy ~180

Bulk-write or clear permission overrides: create expands resources x actions into "resource:action" permission grants for one role, delete clears them, list shows the overrides currently set. Valid expansions: registry:approve, registry:deprecate, team:manage_rbac, team:manage_sso. Requires the team:manage_rbac permission and Enterprise tier (rbac feature).

NameTypeReqDescription
actionstringyesPolicy operation
actionsarrayAction segments, e.g. ["approve"] (required for create/delete)
effectstringPolicy effect (required for create)
namestringDisplay-only label; not persisted
resourcesarrayResource segments, e.g. ["registry"] (required for create/delete)
rolestringRole the policy applies to (required for create/delete)

No output schema declared.

No examples provided.

rbac_manage ~154

Inspect and configure team role permissions: list_roles, get_role, set_role_permission, reset_role_permission. Roles are fixed (owner, admin, member) — owners always hold every permission and cannot be narrowed. Configurable permissions: registry:approve, registry:deprecate, team:manage_rbac, team:manage_sso. Requires the team:manage_rbac permission and Enterprise tier (rbac feature).

NameTypeReqDescription
actionstringyesRBAC permission operation
effectstringOverride effect (required for set_role_permission)
permissionstringPermission to configure (required for set/reset_role_permission)
rolestringRole to inspect or configure (required except for list_roles)

No output schema declared.

No examples provided.

search ~438

[Skillsmith — Discover stage] Search the Skillsmith registry of agent skills (SKILL.md format) — curated, security-scanned, trust-scored skills indexed daily from GitHub. Skillsmith is a registry for sharing, scanning, and tracking agent skills across any MCP-capable runtime. Use this tool for ANY user request to find/search/discover/list skills — e.g. 'search for testing skills', 'find git workflow skills', 'show me devops skills with quality above 80'. Returns ranked installable skills with trust badges, NOT general programming guidance. Results are installable-only by default (pass installable_only:false to also include discovery-only entries that cannot be installed). Filters: query (required), category, trust_tier (verified/curated/community/experimental), min_score, max_risk, safe_only, installable_only, limit, compatibility (IDE/LLM). Matching is keyword-based, not semantic — use a short single-topic query; on empty results, check the response suggestion field for what to try next.

NameTypeReqDescription
categorystringFilter by skill category
compatible_withobjectRank by IDE and/or LLM compatibility (deprioritizes, never excludes)
installable_onlybooleanWhen true (default), return only installable skills — excludes discovery-only registry entries that install_skill cannot resolve. Pass false to opt back in to discovery-only entries.
limitnumberMaximum results to return (default 10). Out-of-range values are clamped, not rejected.
max_risknumberMaximum risk score (0-100, lower is safer)
min_scorenumberMinimum quality score (0-100)
querystringSearch query — matched literally/lexically (not semantically); use a short single-topic term (e.g. 'testing') rather than a multi-concept phrase for best results
safe_onlybooleanOnly show skills that passed security scan
trust_tierstringFilter by trust tier level (verified, curated, community, experimental, unknown)
NameTypeReqDescription
compatibilityDeprioritizednumber
discoveryOnlyHiddennumber
filtersobjectyes
querystringyes
resultsarrayyes
suggestionstring
timingobjectyes
totalnumberyes

No examples provided.

share_skill ~69

Add, remove, or list skills in a team workspace. Requires Team tier license.

NameTypeReqDescription
actionstringyesSharing operation to perform
skillIdstringSkill ID in author/name format (required for add/remove)
workspaceIdstringyesWorkspace ID to share skill with

No output schema declared.

No examples provided.

siem_export ~100

Export audit events for SIEM ingestion (CloudWatch, Splunk, Datadog). Requires Enterprise tier (siem_export feature). SIEM destination is configured via environment variables, not tool arguments.

NameTypeReqDescription
endDatestringISO 8601 end date (defaults to now)
formatstringExport format (default json)
startDatestringISO 8601 start date (defaults to 1h ago)

No output schema declared.

No examples provided.

skill_audit ~87

Check installed skills for known security advisories. Requires Team tier or higher (skill_security_audit feature). The advisory system is in early access — the Skillsmith team publishes advisories as security issues are identified. Run `skillsmith sync` to fetch the latest advisories.

NameTypeReqDescription
skillIdsarraySpecific skill IDs to audit (omit to return all skills with active advisories).

No output schema declared.

No examples provided.

skill_compare ~163

[Skillsmith — Evaluate stage] Compare two Skillsmith-registry skills side-by-side. Use when the user wants to compare/contrast/decide-between two specific skills — e.g. 'compare getsentry/commit and microsoft/playwright-cli', 'which is better, X or Y', 'what's the difference between these two skills'. Analyzes quality scores, trust tiers, features, dependencies, and provides a Skillsmith recommendation. Skillsmith is a registry for sharing, scanning, and tracking agent skills across any MCP-capable runtime.

NameTypeReqDescription
skill_astringyesFirst skill ID to compare (e.g., "getsentry/commit")
skill_bstringyesSecond skill ID to compare (e.g., "microsoft/playwright-cli")

No output schema declared.

No examples provided.

skill_diff ~167

Show a section-level diff between two versions of an installed skill. Returns added, removed, and modified headings along with a change type (major/minor/patch) and update recommendation. Requires Individual tier or higher (version_tracking feature).

NameTypeReqDescription
hasLocalModificationsbooleanWhether the installed skill has local edits
newContentstringyesUpdated SKILL.md content
newRiskScorenumberRisk score of the new version (0–100)
oldContentstringyesPrevious SKILL.md content
oldRiskScorenumberRisk score of the old version (0–100)
skillIdstringyesRegistry skill identifier (e.g. "author/skill-name")
trustTierstringRegistry trust tier

No output schema declared.

No examples provided.

skill_inventory_audit ~225

[Skillsmith — Maintain stage] Audit your installed AI coding clients' skill inventories (Claude Code, Cursor, Copilot, and every other Skillsmith-supported client) — plus Claude Code's own commands, agents, and CLAUDE.md trigger rules — for local namespace collisions. Returns rename + prose-edit suggestions keyed by a fresh `auditId`. Never modifies your skills, commands, agents, or CLAUDE.md — it only writes its own audit artifacts under `~/.skillsmith/audits/<auditId>/`. Feed the returned suggestions into `apply_namespace_rename` / `apply_recommended_edit`.

NameTypeReqDescription
applyExclusionsbooleanApply the configured exclusion list to the scan. Defaults to true.
deepbooleanRun the deep (semantic) collision pass. Defaults to false.
homeDirstringOverride the inventory root. Must resolve under os.homedir() or os.tmpdir(); arbitrary paths are rejected.
projectDirstringProject directory whose CLAUDE.md rules are included in the scan.

No output schema declared.

No examples provided.

skill_outdated ~64

Check installed skills for available updates and dependency satisfaction status. Reads the local manifest, hashes each installed SKILL.md, and compares against the latest registry state. Community tier — no license required.

NameTypeReqDescription
include_depsbooleanInclude dependency satisfaction status (default: true)

No output schema declared.

No examples provided.

skill_pack_audit ~181

Audit a skill pack directory for (a) version drift — bundled SKILL.md versions vs. the Skillsmith registry cache — and (b) trigger-quality issues — generic trigger words in skill names/descriptions and generic pack namespaces that misfire Claude's skill-trigger heuristic (SMI-4124). Response is additive: new triggerQuality and namespaceQuality fields appear when check_trigger_quality is true (default); existing fields (skills, driftCount, etc.) are unchanged. Requires Individual tier or higher.

NameTypeReqDescription
check_trigger_qualitybooleanWhen true (default), also flag generic trigger words and generic pack namespaces with rename suggestions. Set false for version-drift-only audits (legacy response shape — triggerQuality and namespace…
pack_pathstringyesAbsolute path to the skill pack root directory (must contain a skills/ subdirectory).

No output schema declared.

No examples provided.

skill_publish ~167

Prepare a skill for publishing. Validates the skill, generates a checksum, creates a publish manifest, and optionally checks for project-specific references. Can also create a GitHub repository for registry discovery.

NameTypeReqDescription
add_topicbooleanAdd claude-skill topic for registry discovery
check_referencesbooleanRun reference check before publishing (default: true)
create_repobooleanCreate GitHub repository (requires gh CLI — will pre-flight check)
reference_patternsarrayAdditional regex patterns to check (regex strings, max 200 chars each, max 20 patterns)
skill_pathstringyesPath to the skill directory (must contain SKILL.md)
visibilitystringGitHub visibility if creating repo (default: 'public')

No output schema declared.

No examples provided.

skill_recommend ~407

[Skillsmith — Discover stage] Recommend skills from the Skillsmith registry based on the user's project context and currently installed skills, using semantic similarity. Use when the user asks for recommendations or 'what skills should I use' — e.g. 'recommend skills for my React project'. Auto-detects installed skills from ~/.claude/skills/ when not provided. Optional role-based filtering. Returns ranked Skillsmith candidates with description + security, NOT general programming advice. Narrate security per item: absent = never scanned (say so, imply neither safe nor unsafe); passed:null = scanned, no verdict yet; passed:true/false = real verdict — state it, and flag false or high riskScore explicitly rather than recommending install uncritically. Skillsmith is a registry for sharing, scanning, and tracking agent skills across any MCP-capable runtime. candidates_considered:0 does NOT mean a backend problem — check the response's suggestion field for next steps.

NameTypeReqDescription
detect_overlapbooleanEnable overlap detection to filter similar skills (default true)
installable_onlybooleanSMI-5178: When true (default), return only installable skills. Pass false to include discovery-only entries that cannot be installed.
installed_skillsarrayCurrently installed skill IDs (e.g., ["getsentry/commit", "microsoft/playwright-cli"]). If empty, auto-detects from ~/.claude/skills/
limitnumberMaximum recommendations to return (default 5, max 50)
min_similaritynumberMinimum similarity threshold (0-1, default 0.3)
project_contextstringOptional project description for context-aware recommendations (e.g., "React frontend with Jest testing")
rolestringSMI-1631: Filter by skill role (code-quality, testing, documentation, workflow, security, development-partner). Skills matching the role get a +30 score boost.

No output schema declared.

No examples provided.

skill_recover_source ~183

[Skillsmith — Maintain stage] Recover the canonical GitHub source of locally-installed skills (SMI-5407). Read-only — never mutates ~/.skillsmith/manifest.json. Returns { skills, summary } where each skill entry reports confidence (exact/high/medium/low), method, and the recovered source URL. Feed exact/high results into `sklx audit sources --apply` to backfill the manifest.

NameTypeReqDescription
catalogHintbooleanEnable catalog / author hint tier (off by default).
embeddingbooleanEnable embedding tiebreak tier (off by default).
homeDirstringOverride the skills root parent. Must resolve under os.homedir() or os.tmpdir(); arbitrary paths (e.g. /etc) are rejected.
onlyarrayRestrict recovery to these skill directory basenames.

No output schema declared.

No examples provided.

skill_rescan ~93

Re-scan installed skills with the latest security patterns. Detects issues like SSRF instructions, prompt injection, data exfiltration, and other threats that may not have been caught when the skill was originally installed. Run without arguments to scan all installed skills, or specify a skill name to scan one.

NameTypeReqDescription
skillNamestringSpecific skill directory name to rescan (omit to rescan all installed skills).

No output schema declared.

No examples provided.

skill_suggest ~181

Proactively suggest relevant skills based on current context (files, commands, errors, project structure). Counts against your monthly API quota (Community: 100/mo — see www.skillsmith.app/pricing).

NameTypeReqDescription
current_filestringCurrent file being edited (e.g., "src/App.test.tsx")
error_messagestringRecent error message if any
installed_skillsarrayCurrently installed skill IDs (for filtering)
limitnumberMaximum suggestions to return (default 3, max 10)
project_pathstringyesRoot path of the project to analyze
recent_commandsarrayRecent terminal commands (last 5, e.g., ["npm test", "git commit"])
session_idstringSession identifier (optional, for informational purposes — default: "default")

No output schema declared.

No examples provided.

skill_updates ~73

Check installed skills for available updates by comparing locally-recorded content hashes against the current registry state. Requires Individual tier or higher. Returns a list of skills with their installed vs. latest hash and whether an update is available.

NameTypeReqDescription
skillIdsarraySpecific skill IDs to check. Omit to check all tracked skills.

No output schema declared.

No examples provided.

skill_validate ~182

[Skillsmith — Install stage] Validate a SKILL.md file or skill directory against the Skillsmith specification before installing or publishing. Use when the user wants to check/validate a skill's structure — e.g. 'validate my skill at ./my-skill', 'check if this skill is valid', 'use Skillsmith to validate this SKILL.md'. Checks YAML frontmatter, required fields, file structure, and security-pattern signatures. For end-to-end skill authoring including scaffolding/publishing, use the CLI `skillsmith author` commands. Skillsmith is a registry for sharing, scanning, and tracking agent skills across any MCP-capable runtime.

NameTypeReqDescription
skill_pathstringyesPath to SKILL.md file or skill directory containing SKILL.md
strictbooleanEnable strict validation mode (default false). Strict mode treats warnings as errors.
NameTypeReqDescription
errorsarrayyes
metadataobject|nullyes
pathstringyes
timingobjectyes
validbooleanyes

No examples provided.

sso_settings ~49

View current SSO/SAML configuration for your organization. Requires Enterprise tier (sso_saml feature).

NameTypeReqDescription
includeMetadatabooleanInclude full IdP metadata in response (default: false)

No output schema declared.

No examples provided.

team_analytics_dashboard ~51

View team usage analytics: per-user tool usage counts, top tools, and daily trend. Requires Team tier (usage_analytics feature).

NameTypeReqDescription
periodstringTime period (default 30d)

No output schema declared.

No examples provided.

team_usage_report ~57

Generate a weekly/monthly usage summary with period-over-period comparison. Requires Team tier (usage_analytics feature).

NameTypeReqDescription
formatstringReport format (default summary)
periodstringTime period (default 30d)

No output schema declared.

No examples provided.

team_workspace ~78

Manage team workspaces (create, list, get, delete). Requires Team tier license.

NameTypeReqDescription
actionstringyesWorkspace operation to perform
descriptionstringWorkspace description (optional for create)
namestringWorkspace name (required for create)
workspaceIdstringWorkspace ID (required for get/delete)

No output schema declared.

No examples provided.

undo_apply ~228

[Skillsmith — Maintain stage] Undo the most recent apply_namespace_rename / apply_recommended_edit changeset(s) made in THIS server session. Session-scoped: restarting the MCP server clears the undo history — for a namespace rename applied in a PRIOR session, use the durable, cross-session revert instead: apply_namespace_rename({ auditId, collisionId, action: 'revert' }) (SMI-5671). Restores from the apply tool's own pre-mutation backup and refuses (never throws) if the target was modified since the apply, the backup is missing, or the restore target falls outside the confined skill roots. Pass `count` to undo the N most-recent changesets (default 1), or `suggestion_id` to undo one specific changeset — mutually exclusive.

NameTypeReqDescription
countnumberNumber of most-recent session applies to undo, most-recent-first. Defaults to 1. Mutually exclusive with suggestion_id.
suggestion_idstringUndo one specific changeset by its collisionId. Mutually exclusive with count.

No output schema declared.

No examples provided.

uninstall_skill ~265

[Skillsmith — Retire stage] Uninstall an agent skill from the local Claude Code skills directory (~/.claude/skills/) or runtime-equivalent path. Use when the user asks to uninstall/remove/delete a specific skill — e.g. 'uninstall playwright-cli', 'remove getsentry/commit', 'use Skillsmith to delete the testing skill'. Optional `force` flag overrides protection on locally-modified skills. Skillsmith is a registry for sharing, scanning, and tracking agent skills across any MCP-capable runtime.

NameTypeReqDescription
clientstringTarget agent (default: SKILLSMITH_CLIENT env or claude-code)
cwdstringAbsolute path to the calling client's actual project/workspace root, used as the ancestor-walk starting point for workspace scope resolution. Optional.
forcebooleanForce removal even if skill has been modified
scopestringUninstall scope (ADR-139): "workspace" targets the nearest ancestor workspace marker/.git root's skills directory (walked from cwd when provided); defaults to SKILLSMITH_SCOPE env, then the per-clien…
skillNamestringyesName of the skill to uninstall

No output schema declared.

No examples provided.

usage_report ~51

Comprehensive enterprise usage report with all metrics. Requires Enterprise tier (advanced_analytics feature).

NameTypeReqDescription
formatstringReport format (default summary)
periodstringTime period (default 30d)

No output schema declared.

No examples provided.

webhook_configure ~112

Configure webhooks for skill lifecycle events (skill.install, skill.publish, etc.). Webhooks are signed with HMAC-SHA256. Requires Enterprise tier (custom_integrations feature).

NameTypeReqDescription
actionstringyesWebhook operation
descriptionstringWebhook description
eventsarrayEvent types (required for create)
urlstringWebhook URL (required for create)
webhookIdstringWebhook ID (required for get/delete/test/rotate_secret)

No output schema declared.

No examples provided.

Common questions

What is the Skillsmith MCP server?

Skillsmith is an MCP server listed in the public MCP registry as io.github.smith-horn/skillsmith. Share agent skills across your team, scan them for risk, and track what's actually used. This page covers its npm package (@skillsmith/mcp-server).

Is the Skillsmith MCP server safe to use?

Skillsmith scores 88 out of 100 on VerifyMCP. We recorded 2 known advisories against it as of 20 September 2026. It declares no install or post-install scripts. Its build provenance is signed and verified. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

What tools does the Skillsmith MCP server expose?

Skillsmith exposes 43 tools: search, get_skill, install_skill, uninstall_skill, skill_recommend, and 38 more. Their descriptions and schemas cost roughly 7,275 tokens of context every time the server is loaded.

Is the Skillsmith MCP server still maintained?

Skillsmith is still listed as active in the MCP registry. We last reached this channel on 20 September 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

What licence is the Skillsmith MCP server under?

Skillsmith declares the Elastic-2.0 licence, which is not on the OSI-approved list. Read the terms before using it at work, and note this covers the source only, not the cost of any service it calls.