SkillPlus Python SDK

Official Python SDK for the hosted SkillPlus API.

SkillPlus provides security intelligence for AI skills and agent-facing software, helping developers, teams, and platforms scan, understand, and trust AI skills before installation, approval, or integration.

Install

pip install skillplus

Requires Python 3.10+.

Quick start

query gives a binary answer: is there a report, or not.

from skillplus import SkillPlus

client = SkillPlus(api_key="skp_...")

result = client.query("https://www.skills.sh/vercel-labs/skills/find-skills")
# GitHub repositories work the same way: "https://github.com/owner/repo"

if result.status == "found":
    print(result.report.verdict)  # "safe" | "medium" | "high" | "unknown" — UI labels: Safe / Caution / High Risk / Unrated
else:
    print("no report yet", "(scan in progress)" if result.scanning else "")

Prefer verdict over the legacy rating field — it folds historical values onto the current three-tier scale.

The three integration patterns

1. Show a report (frontend page) — pure read, instant:

result = client.query(url)          # found | not_found, never blocks

2. Scan in the background (batch pipelines) — fire-and-forget, or wait:

ack = client.scan(url)              # returns immediately: ack.accepted / ack.scanning
report = client.scan(url, wait=True)   # blocks until the report completes, returns it

3. Query-and-scan (skill detail pages) — if it was never scanned, scan it:

result = client.query(url, scan_if_missing=True)   # triggers a scan, still returns immediately
result = client.query(url, wait=True)              # scans if needed AND waits for the report

wait=True implies scan_if_missing; a failing scan raises SkillPlusError (errors travel on the exception channel, never as a status value).

Client options

client = SkillPlus(
    api_key="skp_...",
    base_url="https://skillplus.xyz",  # staging / self-hosted override
    timeout=30.0,                      # per-request timeout (seconds)
    max_retries=2,                     # retries on 429/502/503, honoring Retry-After
)

Waiting calls accept wait_interval (default 5 s) and wait_timeout (default 600 s).

Context manager

from skillplus import SkillPlus

with SkillPlus(api_key="skp_...") as client:
    ack = client.scan("https://www.skills.sh/vercel-labs/skills/find-skills")
    print(ack.accepted)

Methods

Error handling

from skillplus import SkillPlus, SkillPlusError

client = SkillPlus(api_key="skp_...")

try:
    result = client.query("https://www.skills.sh/vercel-labs/skills/find-skills")
except SkillPlusError as error:
    print(error.status_code)
    print(error.message)

Development

uv sync --extra dev
uv run pytest
uv build