Permission Boundary Explainability Doctor 20260718

Overview

This draft skill converts operator notes into a deterministic PASS / REVIEW / BLOCKED evidence report for the 2026-07-18 Merlin priority. It is a stdlib-first local analyzer that Viper can exercise with clean, risky, missing-input, and sensitive-path fixtures.

This is a draft-only workspace skill. It does not publish, package, deploy, create billing products, mutate cron jobs, edit OpenClaw configuration, or perform runtime remediation.

Merlin Source

• Brief: memory/merlin-build-brief.md dated 2026-07-18.
• Priority signal: governance, plugin permissions, and connector safety remain trust blockers for paid compliance/security users.
• Build target: local draft under skills/permission-boundary-explainability-doctor-20260718/ for downstream Viper testing.

When to Use

• A Merlin brief or operator note identifies this problem class as a paid skill opportunity.
• You need repeatable evidence before Viper scenario testing.
• The operator wants deterministic triage rather than open-ended narrative review.
• You need a safe local script that can be compiled and self-tested in CI-style gates.

Safety

• Stdlib-only Python.
• Local notes or direct text only; hidden and sensitive-looking paths are refused.
• No network calls, subprocess execution, package installation, credential output, runtime repair, deploys, billing changes, cron edits, or OpenClaw config mutation.
• Output includes mutated_files: [] for gate automation.

Main Script

• scripts/permission_boundary_explainability_doctor.py
• Supported modes: --notes, --text, --self-test, --json

Operator Workflow

1. Save non-sensitive notes in a normal text file.

2. Run python3 scripts/permission_boundary_explainability_doctor.py --notes path/to/notes.txt --json.

3. Review status, score, signals_found, risks_found, missing_signals, findings, and recommendations.

4. Add missing evidence or approval gates before handing the draft to Viper.

Viper Scenario Seeds

• Complete evidence notes should PASS.
• Risk-heavy or uncontrolled notes should return BLOCKED.
• Missing notes should return ok=false, status=ERROR, and error=missing_notes.
• Hidden or sensitive-looking paths should be refused and must not be read.

Verification Checklist

• [ ] python3 -m py_compile scripts/permission_boundary_explainability_doctor.py exits 0.
• [ ] python3 scripts/permission_boundary_explainability_doctor.py --self-test --json returns ok=true.
• [ ] Clean fixture resolves to PASS.
• [ ] Risky fixture resolves to BLOCKED.
• [ ] Static/security scans find no execution, remote access, config mutation, cron mutation, deploy, billing, or credential-output path.
• [ ] Draft stays in PENDING_VIPER until downstream review is complete.

Draft Status

PENDING_VIPER — retain this draft state until independent Viper review; the current Iceman gate evidence path is recorded in the daily handoff.