{
  "markdown": "# Persona Plugin for Claude Code\n\nSwitch Claude into expert modes mid-conversation. One command activates a specialist persona — security expert, debugger, architect, and more — shaping Claude's priorities, communication style, and domain focus for the rest of the session.\n\nPersonas are **session-only**. They apply to the current conversation and reset when the session ends.\n\n---\n\n## What it does\n\nThe `/persona` command lets you:\n\n- **Activate** a built-in or custom expert persona for your session\n- **Spawn** a single persona as a sub-agent to handle a specific task\n- **Multi** — run a task across several relevant personas in parallel and get a synthesized result\n- **List** all available personas\n- **View references** a persona uses for its expertise\n- **Create** new personas with a guided template\n- **Edit** existing personas\n- **Delete** personas you no longer need\n- **Deactivate** the current persona and return to default Claude behavior\n\n---\n\n## Prerequisites\n\n- [Claude Code](https://claude.ai/code) installed and authenticated\n\n---\n\n## Installation\n\nClone the repository:\n\n```bash\ngit clone https://github.com/tatertot365/claude-persona-plugin\n```\n\nThen add it as a plugin in Claude Code. The exact CLI flag depends on your Claude Code version — check `claude --help` or your Claude Code settings for the plugin installation option. Point it at the cloned directory.\n\nClaude Code discovers the `/persona` skill automatically from the plugin's `skills/persona/SKILL.md` file.\n\n---\n\n## Quick start\n\n```\n/persona security-expert\n```\n\nClaude confirms activation and immediately responds as a senior application security engineer. Run `/persona off` to return to default behavior.\n\n---\n\n## Command reference\n\n### List available personas\n\n```\n/persona list\n```\n\nOutput:\n\n```\nAvailable Personas\n──────────────────────────────────────────────────────────────────\n  architect          — Principal software architect; thinks in trade-offs, not best practices\n  code-reviewer      — Meticulous reviewer who finds what matters and communicates it clearly\n  debugger           — Methodical, hypothesis-driven bug diagnosis across complex systems\n  security-expert    — Adversarial mindset; looks for how code can be exploited\n  ...\n\nUse /persona <name> to activate · /persona off to deactivate · /persona create to add a new one · /persona edit <name> to edit · /persona delete <name> to delete\n```\n\n---\n\n### Activate a persona\n\n```\n/persona <name>\n```\n\nFuzzy matching is supported — partial names work:\n\n```\n/persona security      → activates security-expert\n/persona debug         → activates debugger\n```\n\nClaude confirms activation and immediately adopts the persona's priorities, expertise, and communication style for every subsequent response in the session.\n\n---\n\n### View a persona's references\n\n```\n/persona ref <name>\n```\n\nLists the reference documents bundled with the persona and when to consult each one.\n\n```\nReferences for security-expert\n──────────────────────\n• owasp-top10.md — consult for web application code audits and security reviews\n• owasp-api-security.md — consult when reviewing API endpoints, authentication flows, or data exposure\n• cwe-quick-reference.md — consult when citing specific vulnerability classes or CWE identifiers\n\nUse /persona ref <name> load <filename> to load a reference into the current session.\n```\n\nTo load a reference file into your session context:\n\n```\n/persona ref security-expert load owasp-top10.md\n```\n\n---\n\n### Spawn a persona as a sub-agent\n\n```\n/persona spawn <name> <task>\n```\n\nExample:\n\n```\n/persona spawn security-expert review this authentication flow for vulnerabilities\n```\n\nLaunches a sub-agent that completes the task entirely through the specified persona's lens, then returns the result. Your session persona is unaffected.\n\nUse `spawn` when you want a one-off expert opinion without switching your session context.\n\n---\n\n### Run a task across multiple personas in parallel\n\n```\n/persona multi <task>\n```\n\nExample:\n\n```\n/persona multi design a REST API for user authentication\n```\n\nClaude selects 2–4 personas most relevant to the task, spawns one sub-agent per persona simultaneously, and presents each result in a labeled section. A **Synthesis** section follows — key points of agreement and disagreement across all perspectives.\n\nExample personas Claude might select for the above task: `architect`, `security-expert`, `senior-engineer`.\n\nYour session persona is unaffected.\n\n---\n\n### Deactivate the current persona\n\n```\n/persona off\n```\n\nReturns Claude to default behavior. Also accepts `/persona deactivate` and `/persona reset`.\n\n---\n\n### Create a new persona\n\n```\n/persona create\n```\n\nClaude prompts you to fill out a template:\n\n```\nName:       <lowercase-hyphenated, e.g. api-architect>\nRole:       <one sentence describing the expert role>\nTone:       <communication style, e.g. direct, educational, rigorous, warm>\nPriorities: <2–3 things this expert always optimizes for>\nExpertise:  <key domains, tools, or methodologies>\nPitfalls:   <common mistakes or anti-patterns this expert watches for (optional)>\n```\n\nClaude generates a `.md` file from your input and saves it to `skills/persona/personas/`. It then asks whether to activate the new persona immediately.\n\n---\n\n### Edit a persona\n\n```\n/persona edit <name>\n```\n\nClaude displays the current file and asks what you'd like to change. After you describe the changes, it rewrites the file and confirms what was updated.\n\n---\n\n### Delete a persona\n\n```\n/persona delete <name>\n```\n\nClaude asks for confirmation before deleting. This action cannot be undone.\n\n---\n\n### Get help\n\n```\n/persona help\n```\n\nPrints the full command reference. Running `/persona` with no arguments shows the same output.\n\n---\n\n## Session persona vs. sub-agents\n\nThese two modes operate independently and can run at the same time:\n\n| | Session (`/persona <name>`) | Sub-agent (`spawn` / `multi`) |\n|---|---|---|\n| **Scope** | Shapes every response in the session | Handles one specific task |\n| **Persistence** | Active until `/persona off` or session ends | Completes and exits |\n| **Combinable** | Yes | Yes |\n\nExample: activate `security-expert` as your session persona, then run `/persona spawn architect design a caching layer` — both operate simultaneously without interfering.\n\n---\n\n## Built-in personas\n\n| Persona | Focus |\n|---|---|\n| `architect` | System design, trade-offs, scalability |\n| `code-reviewer` | Code quality, correctness, clear feedback |\n| `data-scientist` | Statistical analysis, ML, actionable insights |\n| `debugger` | Root-cause analysis across complex systems |\n| `graphic-designer` | Visual design, UI/UX, design systems |\n| `legal-reviewer` | Tech contracts, compliance, risk flags |\n| `product-manager` | User needs, prioritization, stakeholder alignment |\n| `product-researcher` | UX research, user behavior, qualitative insights |\n| `security-expert` | Adversarial security, threat modeling, OWASP |\n| `senior-engineer` | Pragmatic, production-hardened engineering |\n| `tech-writer` | Clear documentation, developer guides, API references |\n\n---\n\n## Adding your own personas\n\nPersonas are plain Markdown files in `skills/persona/personas/`. Create one manually or use `/persona create`.\n\nEach file follows this structure:\n\n```markdown\n# Name\n\nOne paragraph describing who this expert is and their default mindset.\n\n**Priorities:**\n- Priority 1\n- Priority 2\n\n**How you work:**\n- Behavioral guideline 1\n- Behavioral guideline 2\n\n**Communication style:**\n- Style guideline 1\n\n**Domain expertise:**\n- Domain or tool\n\n**Pitfalls you watch for:**\n- Common mistake to avoid\n```\n\nSave the file as `<name>.md` in the `personas/` directory. It appears in `/persona list` immediately.\n\nTo give a persona reference documents (files it can consult during a session), add a `**References:**` section to the persona file and place the reference files in `skills/persona/references/<name>/`. Shared references that apply to multiple personas go in `skills/persona/references/shared/`.\n\n```markdown\n**References:**\n- `references/<name>/my-reference.md` — consult when ...\n- `references/shared/cwe-quick-reference.md` — consult when citing CWE identifiers\n```\n\n---\n\n## How it works\n\nThe skill is defined in `skills/persona/SKILL.md`. When you run `/persona <arguments>`, Claude Code loads that file and executes its instructions using the `Read`, `Write`, `Glob`, and `Bash` tools.\n\nPersona files are read at activation time. Claude ingests the full file and applies it as behavioral context for the rest of the session. References are loaded on demand — not upfront — based on when the task warrants them.\n\nNo model fine-tuning or external APIs are involved.\n\n---\n\n## Scripts\n\nSome personas can run scripts before applying their reasoning. Scripts produce deterministic data — file lists, complexity scores, matched patterns — that give the persona a concrete foundation to work from.\n\nScripts do not make decisions. They surface candidates for human review and persona interpretation.\n\n### How scripts fit into a workflow\n\n1. Run the relevant script via the Bash tool to collect data.\n2. The active persona interprets the results and produces findings.\n3. Review the script output before acting on any finding — scripts flag candidates, not confirmed issues.\n\n### Available scripts\n\n| Script | Used by | When to run |\n|---|---|---|\n| `scan-secrets.sh [path]` | `security-expert` | Before any security review — surfaces hardcoded credentials, tokens, and API keys |\n| `find-injection-sinks.py [path]` | `security-expert` | When reviewing code for injection vulnerabilities — surfaces candidates for human verification |\n| `review-diff.sh [ref_or_file]` | `code-reviewer` | At the start of a code review — generates a structured inventory of files, functions, exports, TODOs, and risk flags |\n| `find-unhandled-errors.py [path]` | `code-reviewer` | When checking error handling coverage across Go, Python, JS/TS, Java, and Rust |\n| `extract-stack-trace.sh <file\\|->` | `debugger` | When given logs or crash output — extracts, deduplicates, and ranks error types by frequency |\n| `git-diff-working-broken.sh <good> <bad>` | `debugger` | When diagnosing a regression — produces a structured diff between a working and broken ref, with a git bisect hint |\n| `dep-graph.sh [path]` | `architect` | When assessing coupling or planning a refactor — maps module import relationships and coupling hot spots |\n| `count-complexity.py [path] [--threshold N]` | `code-reviewer`, `architect`, `debugger` | When identifying high-risk functions — cyclomatic complexity per function, ranked by severity |\n\nScripts live in `skills/persona/scripts/`. Persona files reference their associated scripts in a `**Scripts:**` section so they know when and how to use them.\n\n---\n\n## File structure\n\n```\nclaude-persona-plugin/\n├── .claude-plugin/\n│   └── plugin.json              # Plugin metadata (hidden directory)\n└── skills/\n    └── persona/\n        ├── SKILL.md              # Skill definition (entry point)\n        ├── personas/\n        │   ├── architect.md\n        │   ├── code-reviewer.md\n        │   ├── data-scientist.md\n        │   ├── debugger.md\n        │   ├── graphic-designer.md\n        │   ├── legal-reviewer.md\n        │   ├── product-manager.md\n        │   ├── product-researcher.md\n        │   ├── security-expert.md\n        │   ├── senior-engineer.md\n        │   └── tech-writer.md\n        ├── scripts/\n        │   ├── scan-secrets.sh\n        │   ├── find-injection-sinks.py\n        │   ├── review-diff.sh\n        │   ├── find-unhandled-errors.py\n        │   ├── extract-stack-trace.sh\n        │   ├── git-diff-working-broken.sh\n        │   ├── dep-graph.sh\n        │   └── count-complexity.py\n        └── references/\n            ├── shared/           # Reference files available to all personas\n            │   └── cwe-quick-reference.md\n            ├── architect/\n            │   ├── adr-template.md\n            │   └── system-design-patterns.md\n            ├── security-expert/\n            │   ├── owasp-top10.md\n            │   └── owasp-api-security.md\n            └── ...               # One directory per persona (if it has references)\n```\n",
  "bytes": 12171,
  "sha": "6dcaa757fc54faba74d31bc2557e940278cd403104dc1229fc1b2f7660919c6f",
  "repo_slug": "tatertot365/claude-persona-plugin",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_tatertot365_claude_persona_plugin_person_3f7d29ce/readme"
}