Skip to content

Rule Mechanics

The category pages describe what rules claim. This page describes the engine that evaluates them: what a rule sees, how a match becomes weight, how rules are configured, and how identifiers are allocated. Each subject has one owner; this page is the landing point.

The evaluation path

  1. Parse and resolve. The PKGBUILD is read into a typed document and, where static resolution allows, variable references are expanded. What cannot be resolved is recorded as a boundary, never guessed. See Static Analysis and Resolution.
  2. Match. Regex and structural rules run against the chosen surface; heuristic, graph and history rules run in code because they need context a single line cannot see. See Rule Matching.
  3. Score. Every finding contributes its severity weight. Priors and maturity-gated novelty add alongside the rules. See Rule Scoring.
  4. Report. Findings, coverage states and the verdict are rendered deterministically. See Report Schema.

The five mechanics pages

Subject Owner Covers
Scoring Rule Scoring Severity weights, score calculation, FATAL, review thresholds
Matching Rule Matching match_target, scope, resolved versus raw-line matching, limitations
Configuration Rule Configuration rules.toml fields, config.toml overrides, structural rules, enablement
Identifiers Rule Identifiers Reserved ranges, allocation, retired and skipped IDs
Series Rule Series Mechanism definitions and series-specific detection behaviour

Which file controls what

File Controls Reference
Shipped rules.toml Regex patterns, targets, scopes, severities Rule Configuration
config.toml Per-rule overrides, enablement, thresholds, profiles Configuration Reference
src/trustsight/categories.py The closed category set and the page each category owns Rule Categories
src/trustsight/rule_id_history.py Retired and renamed identifiers Rule Identifiers
scripts/build_rules_index.py Generated category tables and the full catalog Rules Overview

Where to go next