Understanding security scores
How Foo Guard scores and grades agent configurations.
Page text and source link for your AI assistant. Nothing is sent automatically.
Foo Guard computes a deterministic score from triggered findings and their severities.
A security score of 100 means no scored findings; 0 means the maximum scored risk. This is not proof that an agent is safe. Each finding deducts severity-weighted points from 100. Grades and policy decisions are unchanged.
API results identify this scale as security-v2. CLI baseline version 2 uses this scale; the updated CLI converts legacy version 1 baselines when comparing them. CLI JSON output is version 2. Update older CLI installations before using the new scale. During the transition, requests explicitly identifying an older CLI receive the legacy risk scale so their existing CI comparisons remain correct.
Grades
Grades summarize overall posture on a familiar A–F scale. Lower scores indicate more severe or numerous misconfigurations.
Important notes
- Scores are computed by the rules engine, not by AI
- The same configuration always produces the same score
- Team GitHub scans and repository policy may apply additional pass/fail thresholds