skill Proprietary (Paramus) — guidance only; execution performed by Paramus tools under their own license gating.
Triage a process or chemical inventory for safety hazards by screening reactive-chemistry incompatibilities, flagging hazardous properties (flammability, toxicity, thermal instability), and summarizing key risks using Paramus data and property tools.When to use: a user asks “what are the safety risks of this process/mixture?”.Multi-tool WORKFLOW with process-safety judgment.

Paramus — Process Safety Triage

Overview

Screen an inventory/process for hazardous properties and reactive incompatibilities, then summarize the dominant risks. The value is the fused hazard picture + safety judgment, not one property lookup.

Guidance vs execution: hazard properties and compatibility come from tested Paramus data/tools. Do not assert safety conclusions from model memory — this is a domain where errors are dangerous.

When to Use

  • Early hazard screening of a process, mixture, or chemical inventory
  • Flagging incompatible-storage / reactive-hazard combinations Do not use for: formal HAZOP/LOPA or regulatory safety cases (those require qualified human review).

Workflow

Chemical inventory / process streams + conditions
  ↓ 1. Resolve components            → identifiers + hazard data lookup
  ↓ 2. Hazardous-property flags       → flammability, toxicity, thermal instability, corrosivity
  ↓ 3. Reactive incompatibility check → pairwise incompatibility matrix
  ↓ 4. Condition-aware weighting      → T/P amplifying a hazard
  ↓ 5. Risk summary & report          → ranked hazards + incompatibilities + provenance + disclaimer

Procedure

  1. Discover tools via search/get_schema ("hazard properties", "chemical compatibility").
  2. Resolve components; retrieve hazard data with provenance.
  3. Flag hazardous properties per component.
  4. Screen pairwise reactive incompatibilities across the inventory.
  5. Weight by process conditions; report ranked risks, incompatibilities, and a clear scope disclaimer.

Domain Judgment

  • This is a screen, not a safety assessment — state prominently that qualified human review is required.
  • Absence of a flag ≠ safe; state data-coverage limits explicitly.
  • Incompatibility hits are high-priority — surface them first, never bury them in a score.

Fallbacks

  • Hazard data missing for a component → mark it "unassessed"; never imply it is safe.
  • Endpoint unreachable → stop and report; do not infer hazards from local heuristics.