Skills Graph
Store reusable Agent Skills in Memgraph and track which sessions and subagents use them.
Part of Context Graph. For automatic capture in a coding agent, use the Context Graph installer. The Python API below lets you manage skill definitions directly.
Installation
pip install skills-graph
# With the runtime connector:
pip install 'skills-graph[memgraph-context-graph]'
Requires a running Memgraph, defaulting to bolt://localhost:7687. Direct Python use reads MEMGRAPH_URL, MEMGRAPH_USER, and MEMGRAPH_PASSWORD; hook integrations read the Context Graph config file.
Quick start
from skills_graph import SkillGraph, Skill
sg = SkillGraph()
sg.setup()
sg.add_skill(
Skill(
name="memgraph-cypher",
description="Writing Cypher queries for Memgraph",
content="# Cypher for Memgraph\n\nUse MATCH, CREATE, MERGE ...",
)
)
skill = sg.get_skill("memgraph-cypher")
sg.search_by_name("memgraph")
all_skills = sg.list_skills()
sg.update_skill("memgraph-cypher", content="updated content")
sg.delete_skill("memgraph-cypher")
Skill names use 1–64 lowercase letters, digits, or hyphens, with no leading, trailing, or consecutive hyphens. Invalid names raise SkillValidationError.
Track usage automatically
Wire SkillGraphConnector into an AgentLink and skill usage is recorded automatically from any runtime adapter's event stream (Claude Code, Codex, OpenAI SDK):
from skills_graph import SkillGraph
from skills_graph.connector import SkillGraphConnector
from memgraph_context_graph import AgentLink
link = AgentLink()
skills = SkillGraph()
skills.setup()
link.add_connector(SkillGraphConnector(skills))
For SDK application wiring, see the SDK quick starts.
The connector detects skill usage from TOOL_START/TOOL_END/MESSAGE events — direct tool names like get_skill, MCP-style names like mcp__skills__get_skill, Claude Code's Skill tool, and reads of an Agent Skills definition (.../skills/<name>/SKILL.md) by a file-read tool or a read program such as cat or sed, which covers runtimes where using a skill appears as a file read rather than a dedicated tool call. A SKILL.md that is only mentioned, staged, searched or written is not usage, and neither is other Markdown with skill-like frontmatter. Search/list results that surface skill objects are also recorded through USED_SKILL.
Graph model
(:Skill {name, description, content, license, compatibility,
allowed_tools, metadata, created_at, updated_at, source_path?})
(:Skill)-[:DEPENDS_ON]->(:Skill)
(:Session|:Agent)-[:USED_SKILL {first_access, last_access, access_count, actions}]->(:Skill)
USED_SKILL starts at the subagent's (:Agent) when actions-graph has created it, otherwise at a MERGE (:Session {session_id}) — skills-graph MERGEs the shared Session node it does not own (see the Context Graph map). source_path is set only when a skill is created by observing a local SKILL.md read.
Dependencies between stored skills use sg.add_dependency(skill_name, depends_on) and sg.get_dependencies(skill_name).
Development
From the repository root:
./scripts/dev-memgraph.sh test skills-graph
Metadata
Release files for skills-graph 0.4.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| skills_graph-0.4.0.tar.gz | 24.6 kB | Without provenance | Details for skills_graph-0.4.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| skills_graph-0.4.0-py3-none-any.whl | Python 3 | none | any | Details for skills_graph-0.4.0-py3-none-any.whl |
Total release size: 39.0 kB
Release files / skills_graph-0.4.0.tar.gz
| Download URL | skills_graph-0.4.0.tar.gz |
|---|---|
| Size | 24.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
37b91b84a0226b5fd51ce98b06c55b7fa478df2e4698f8fd7b6af6c18330c90d
|
|
BLAKE2b-256 checksum How to use checksums |
62ff561b3921677e693a14edf2f7f1f2e18b1cea5e731ae6ac2d05d20bc8f7a2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.6 {"installer":{"name":"uv","version":"0.13.6","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / skills_graph-0.4.0-py3-none-any.whl
| Download URL | skills_graph-0.4.0-py3-none-any.whl |
|---|---|
| Size | 14.3 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
a421018c699c19aae9f6f82efd6bbac511fed9e54e9f8ecd7590ade9bb4e41fb
|
|
BLAKE2b-256 checksum How to use checksums |
ef10d68c91c8eb78f018499b5242ee9482ff6cc70cd12f21e418ec58e52a0cfa
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.6 {"installer":{"name":"uv","version":"0.13.6","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Security for skills-graph 0.4.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.