pre-reasoning

name	pre-reasoning
description	Use the Pre-Reasoning Engine to ground autonomous work with structural traces, derived assumptions, and root blockers before answering, then pulse-check the response before stopping.

Pre-Reasoning Skill

Install

pip install pre-reasoning

For a local repo checkout:

pip install -e .

Ground Before Answering

from pre_reasoning import analyze

problem = """A depends on B. B depends on C.
The CTO wants microservices, but the senior dev warns about complexity."""

trace = analyze(problem)["trace"]
print(trace)

The trace includes direct structural blocks plus derived assumptions from the built-in closure expert when transitive dependencies are present. Use it to order the response:

• resolve root blockers first
• follow the unlock sequence
• account for derived assumptions without treating them as direct user claims
• keep parallel work separate
• call out conflicts directly

Pulse Before Stopping

from pre_reasoning import pulse

check = pulse(problem, draft_response)
print(check)

If status is CONTINUE, revise the response to address the listed gaps, then call pulse() again. If status is COMPLETE, the response has addressed the detected root blockers.

Bundled 13.7M trainable parameter model weights are used automatically. Set PRE_REASONING_CHECKPOINT=/path/to/weights.safetensors only when testing another converted weights file.