posturekit 0.1.0

Ingest large, messy data-security exports (.xlsx/.csv) into DuckDB and score them for risk. A model-agnostic DSPM kernel — bring your own data and risk matrix.

posturekit

Ingest large, messy data-security exports and score them for risk — a model-agnostic DSPM kernel. Bring your own data and your own risk matrix.

Two building blocks, usable together or apart:

• ingest — load big, ugly tabular exports (.xlsx / .csv) into DuckDB in constant memory. Survives the things that break naive readers: a multi-GB inner sheet XML with a bogus <dimension>, inline strings, a report banner above the header, and duplicate column names in denormalised exports. Header rows are found by anchor column, so you don't hardcode a row number.
• scoring — a pure, config-driven engine that turns per-resource signals (exposure × sensitive-data × retention) into a risk score, level, SLA and remediation action. No I/O, no model calls — fully unit-testable and reusable across any resource type (mailboxes, folders, sites, drives, shares).

Install

pip install posturekit            # scoring engine only (zero deps)
pip install "posturekit[ingest]"  # + DuckDB / pyarrow / lxml ingest stack

Score resources

from posturekit import score_resource

profile = score_resource({
    "exposure_level": "anyone",
    "classifications": ["credentials", "pii"],
    "sensitive_hit_count": 4200,
    "age_years": 9,
})
# profile["risk_level"]            -> "Critical"
# profile["remediation_priority"] -> "P1"
# profile["sla_days"]             -> 7
# profile["recommended_action"]   -> "Permission Removal"
# profile["risk_combinations"]    -> ["sensitive_stale_and_broad", ...]

The whole model — weights, exposure scale, sensitive-class ceilings, retention buckets, combination boosts, score→level bands, SLAs — lives in one CONFIG dict, and every function takes a config= override. Drop in your own risk matrix without touching the code:

score_resource(signals, config={"bands": [
    {"min_score": 70, "level": "Critical", "priority": "P1", "sla_days": 3},
    {"min_score": 0,  "level": "Low",      "priority": "P4", "sla_days": 180},
]})

Ingest an export

import duckdb
from posturekit import ingest_table

con = duckdb.connect("posture.duckdb")
ingest_table("Resources.xlsx", "resources", con, header_anchor="Path")
ingest_table("Permissions.csv", "permissions", con, header_anchor="Path")

Every column lands as VARCHAR (lossless — cast/derive at query time). The .xlsx and .csv paths produce an identical schema for the same export, so your downstream SQL doesn't care which format the data arrived in. For an export whose layout you know exactly, pass header_row=12 instead of header_anchor.

End to end

from posturekit import score_batch, summarize

cur = con.execute('SELECT * FROM "resources"')
cols = [c[0] for c in cur.description]
rows = [dict(zip(cols, r)) for r in cur.fetchall()]

scored = score_batch(to_signals(r) for r in rows)   # you map columns -> signals
print(summarize(scored))
# {'total': ..., 'by_level': {'Critical': ...}, 'by_combination': {...}}

See examples/demo.py for a runnable build→ingest→score→summarize.

License

MIT