Permission Doctor Skill

Read-only permission analysis for host-platform settings. Detects redundant permissions, security anti-patterns, and validates permission syntax without making changes.

Enforcement

Execution mode: Run scripts exactly as documented; present analysis results without modifying files.

Prohibited actions:

• Do not modify any settings files; this skill is strictly read-only
• Do not invent script arguments not listed in the operations table
• Do not skip anti-pattern detection when analyzing settings

Constraints:

• Platform-neutral audits go through python3 .plan/execute-script.py plan-marshall:platform-runtime:platform_runtime permission analyze --checks {checks} --scope {scope} — the runtime resolves the active platform's settings path and load. This is the preferred entry point.
• The individual permission_doctor detection scripts run through python3 .plan/execute-script.py plan-marshall:tools-permission-doctor:permission_doctor {command} {args}, addressed by --scope (never a literal settings path).
• Use tools-permission-fix for any write operations.
• User-approved permissions must be excluded from suspicious reports.
• Do not hardcode a platform settings-file path (no ~/.claude/settings.json). Address the host platform by --scope; the runtime layer resolves the settings location for the active platform.

What This Skill Provides

Permission Validation Standards

• Syntax validation patterns for all permission types
• Path format validation rules
• Duplicate detection algorithms
• Permission categorization logic

Architecture Patterns

• Global vs Local permission separation
• Universal git access patterns
• Project-specific permission patterns
• Skill and tool permission organization

Security Anti-Patterns

• Suspicious permission detection patterns
• Critical system directory checks
• Dangerous command patterns
• Overly broad wildcard detection

When to Activate This Skill

Activate when:

• Validating permission syntax
• Detecting security anti-patterns
• Understanding global/local architecture
• Analyzing permission issues without making changes

Operations

Operation: detect-redundant

Detect permissions in local settings that duplicate global settings.

Script: permission_doctor.py detect-redundant

Input:

python3 .plan/execute-script.py plan-marshall:tools-permission-doctor:permission_doctor detect-redundant \
  --global-settings {global_path} \
  --local-settings {local_path}

Output (TOON):

redundant[1]{permission,reason,type}:
Bash(git:*)	Exact duplicate	exact_duplicate
marketplace_in_local[1]{permission,reason,type}:
Skill(pm-dev-builder:*)	Should be in global	marketplace_permission
summary:
  redundant_count: 1
  marketplace_in_local_count: 1

Usage: Call before fixing to identify redundancies between global and local settings.

---

Operation: detect-suspicious

Detect permissions matching anti-patterns (security risks).

Script: permission_doctor.py detect-suspicious

Input:

python3 .plan/execute-script.py plan-marshall:tools-permission-doctor:permission_doctor detect-suspicious \
  --settings {settings_path} \
  [--approved-file {run_config_path}]

Output (TOON):

suspicious[1]{permission,reason,severity}:
Write(/tmp/**)	System temp access	medium
already_approved[1]:
- Bash(sudo:*)
summary:
  total_suspicious: 1
  by_severity:
    high: 0
    medium: 1
    low: 0

Usage: Call to identify security anti-patterns. User-approved permissions are excluded.

---

Operation: detect-missing-project-step-permissions

Detect project:{skill} step references in marshal.json that lack matching Skill({skill}) allow rules in settings.

Script: permission_doctor.py detect-missing-project-step-permissions

Input:

python3 .plan/execute-script.py plan-marshall:tools-permission-doctor:permission_doctor detect-missing-project-step-permissions \
  --marshal {marshal_path} \
  (--settings {settings_path} | --scope project|global)

Scan scope: plan.phase-5-execute.steps and plan.phase-6-finalize.steps. Entries starting with project: are enumerated; the substring after project: is matched against permissions.allow as either exact Skill({skill}) or covering wildcard Skill({skill}:*).

Output (TOON):

missing[1]{skill,step,phase}:
finalize-step-plugin-doctor	project:finalize-step-plugin-doctor	phase-6-finalize
present[1]{skill,step,phase,covered_by}:
sync-plugin-cache	project:sync-plugin-cache	phase-6-finalize	Skill(sync-plugin-cache)
summary:
  missing_count: 1
  present_count: 1
  project_steps_checked: 2

Usage: Run during health check and after set-steps configuration to surface missing Skill() allow rules. Pair with tools-permission-fix:apply-project-step-permissions to auto-add missing entries.

---

Operation: analyze (platform-routed)

High-level, platform-neutral analysis. The runtime resolves the active platform's settings, runs the requested checks, and consolidates results — no settings path is named in the body.

Input:

python3 .plan/execute-script.py plan-marshall:platform-runtime:platform_runtime permission analyze \
  --checks redundant,suspicious,missing-steps \
  --scope both \
  [--marshal .plan/marshal.json]

(--marshal is required only when the missing-steps check is included.)

Output (TOON):

status: success
scope: both
checks_run[3]:
- missing-steps
- redundant
- suspicious
total_findings: 5

On a platform with no validated permission backend (e.g. OpenCode), the op returns an honest no-op with a reason and alternative instead of a fabricated finding set.

Usage: Preferred entry point for permission analysis. Consolidates the detection results across checks for the active platform.

Scripts

Script	Subcommand	Purpose
permission_doctor.py	detect-redundant	Detects redundant permissions between global/local
permission_doctor.py	detect-suspicious	Detects security anti-patterns in permissions
permission_doctor.py	detect-missing-project-step-permissions	Detects project:{skill} steps in marshal.json without matching Skill() allow rules
permission_common.py	(library)	Shared utilities for settings loading and path resolution (also used by tools-permission-fix)

Standards Organization

• standards/permission-validation-standards.md - Validation patterns, syntax rules, categorization
• standards/permission-architecture.md - Global/Local separation, universal access patterns
• standards/permission-anti-patterns.md - Security patterns, suspicious permission detection

Non-Prompting Requirements

This skill is designed to run without user prompts. Required permissions:

Script Execution:

• Bash(python3 .plan/execute-script.py *) - Script execution via executor

Settings access is performed inside the runtime / script layer, which resolves the active platform's settings location — the skill body reads no settings file directly.

Ensuring Non-Prompting:

• All operations are read-only analysis
• No file modifications performed
• Script invocation uses executor pattern

Canonical invocations

The canonical argparse surface for permission_doctor.py. The plugin-doctor analyzer (_analyze_manage_invocation.py) reads this section as source-of-truth for the manage-invocation-invalid and missing-canonical-block rules. Consuming docs xref this section by name instead of restating the command inline. See `pm-plugin-development:plugin-script-architecture` cross-skill-integration.md § "Script invocation in documentation".

detect-redundant

python3 .plan/execute-script.py plan-marshall:tools-permission-doctor:permission_doctor detect-redundant \
  (--scope both | --global-settings GLOBAL_SETTINGS) [--local-settings LOCAL_SETTINGS]

--scope and --global-settings are mutually exclusive; --global-settings requires --local-settings.

detect-suspicious

python3 .plan/execute-script.py plan-marshall:tools-permission-doctor:permission_doctor detect-suspicious \
  (--settings SETTINGS | --scope {global,project}) [--approved-file APPROVED_FILE]

--settings and --scope are mutually exclusive.

detect-missing-project-step-permissions

python3 .plan/execute-script.py plan-marshall:tools-permission-doctor:permission_doctor detect-missing-project-step-permissions \
  --marshal MARSHAL (--settings SETTINGS | --scope {global,project})

--settings and --scope are mutually exclusive.

Critical Rules

Read-Only:

• This skill NEVER modifies files
• All operations are analysis and reporting only
• Use tools-permission-fix skill for write operations

Anti-Pattern Detection:

• Uses 24 suspicious patterns from standards/permission-anti-patterns.md
• Severity scoring: high, medium, low
• User-approved permissions are excluded from reports

Part of: plan-marshall-core bundle