/trust-appraise

Runs a non-mutating Trust Magnitude dry run. Works in two modes:

1. Registry node mode (--skill) — appraise an already-curated skill using its live evidence from registry/nodes/. Use this during L4 human review to surface real TM and per-row artifact scores before signing off.
2. Suite proposal mode (--repo) — appraise a proposed suite from live GitHub signals before curation. This is the pre-curation companion to /trust-appraise-all.

Usage

# Appraise one or more curated registry nodes (use during L4 human review)
PYTHONPATH=src python3 scripts/trust_appraise.py \
  --skill rico-favor/implement-with-discernment \
  --skill caioribeiroclw-pixel/evidence-attestation

# Appraise a proposed suite (pre-curation)
PYTHONPATH=src python3 scripts/trust_appraise.py \
  --repo gsd-build/get-shit-done \
  --components 5 \
  --evidence-path docs/INVENTORY.md

# Default suite appraisal set
PYTHONPATH=src python3 scripts/trust_appraise.py

# Machine-readable output
PYTHONPATH=src python3 scripts/trust_appraise.py --skill foo/bar --json

What it scores

Registry node mode (--skill): reads registry/nodes/ directly and calls the same computeTrustMagnitude + computeRowArtifactScores used by the live registry. Shows per-row artifact scores so you can see exactly which evidence entries are contributing TM and which are scoring 0.

Suite proposal mode (--repo): builds a temporary skill object combining live GitHub signals:

• github-stars-own scored under logarithmic diminishing-returns adoption (min(175.0, 35.0 * log10(stars/10.0)) capped at 175 TM; #1705), with skillCountInRepo so mothership discount applies.
• repo-own from live GitHub contributor and contribution counts.
• fusion-recipe from the proposed curated component count — structural only, contributing 0 TM. It feeds the informational Fusion Score, never Trust Magnitude (META.md §2.1e).
• Component repository evidence is bounded to 50.0 TM per suite component to prevent sub-suite cap evasion. Grade S (5★) requires an independent witness row (benchmark-result, verifier-attestation, or Grade A peer-review) and cannot be reached by repository stars alone.

Important caveat

A high dry-run score is not promotion approval. Use this skill to surface evidence quality and TM bias before L4 human review, not to bypass review. In particular: arxiv entries with 0 citations score 0 TM regardless of trustNumber; social-signal entries require verifiable view counts — fabricated numbers directly inflate scores.