Skill Vetting Buyer Trust Report 20260723

Overview

Draft for the matching 2026-07-23 Merlin priority. The stdlib-first analyzer checks third-party skill provenance, permission scope, suspicious-code review, compatibility, sandbox, rollback, and reviewer evidence.

It is advisory and read-only. It performs no installation, execution, repair, publication, packaging, deployment, billing, configuration, scheduler, service-control, deletion, or mutation action and does not certify a downstream state.

When to Use

• Before acting on non-sensitive notes about third-party skill provenance, permission scope, suspicious-code review, compatibility, sandbox, rollback, and reviewer evidence.
• When a deterministic evidence-gap report is needed.
• For Viper clean, risky, missing, guarded-path, and oversized-input scenarios.

Safety Contract

• Reads only direct --text or one explicitly supplied local notes file.
• Refuses hidden or sensitive-looking paths, symlinks, non-files, and inputs over 200,000 characters.
• Uses standard-library modules only; every report includes mutated_files: [].
• Analyzer PASS is not independent approval.

Main Script

scripts/skill_vetting_buyer_trust_report.py supports --notes, --text, --self-test, and --json.

Operator Workflow

1. Prepare non-sensitive evidence notes with no private values or personal data.

2. Run python3 scripts/skill_vetting_buyer_trust_report.py --notes notes.txt --json; completion means structured JSON is returned without input mutation.

3. Assign each finding or gap to an owner or explicit deferral.

4. Hand the unchanged draft and evidence to Viper while retaining PENDING_VIPER.

Result Semantics

• PASS: at least nine expected signals and no risk signal; analyzer result only.
• REVIEW: partial evidence or one bounded concern.
• BLOCKED: multiple risk signals or insufficient evidence.
• ERROR: missing, guarded, oversized, or unreadable input.

Common Pitfalls

1. Treating local PASS as downstream approval.

2. Supplying private values instead of evidence references.

3. Using recommendations as an execution engine.

4. Omitting rollback and reviewer evidence.

Verification Checklist

• [ ] python3 -m py_compile scripts/skill_vetting_buyer_trust_report.py exits 0.
• [ ] Self-test returns clean PASS, risky BLOCKED, and no mutations.
• [ ] Path/size guards and targeted security/static checks pass.
• [ ] Draft remains PENDING_VIPER.

Draft Status

PENDING_VIPER — local build evidence is not an independent downstream decision.