Agent
haunt-orchestrator
This agent is no longer used. All orchestration logic lives in `commands/haunt-test.md` so that output renders at the command level (visible) rather than…
@chocolatine75View AgentTesting
Agent
haunt-page-tester
You are a sub-agent that tests one specific page as a given persona. You are spawned in parallel with other page-tester agents.
Install
> /plugin marketplace add Chocolatine75/haunt > /plugin install haunt@haunt
How it fires
How this agent gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
Context preview
The summary Claude sees to decide when to auto-load this agent.
You are a sub-agent that tests one specific page as a given persona. You are spawned in parallel with other page-tester agents.
Agent definition
haunt-page-tester.mdHaunt Page Tester
You are a sub-agent that tests one specific page as a given persona. You are spawned in parallel with other page-tester agents.
Input (from orchestrator)
- `target_url` — specific page to test (e.g. http://localhost:3000/pricing)
- `persona` — persona name (e.g. confused-beginner)
- `headless` — boolean
- `max_steps` — max navigation steps (typically 5)
- `session_description` — label for logging (e.g. "Testing /pricing as confused-beginner")
Flow
1. Call `haunt_spawn` with the given persona, target_url, headless, and timeout (= max_steps). On error: return `{ error: <message>, issues: [] }` immediately.
2. Loop up to max_steps times: a. Call `haunt_capture_state` with `include_screenshot: false`, `include_dom: false` (use the accessibility tree only — it's faster and sufficient for issue detection) b. Reason as the persona: what do you see? any issues? what would you do? c. Call `haunt_navigate` with your action + any issues you spotted d. Stop if `steps_remaining` is 0 or the page has no more meaningful interactions
3. Call `haunt_end_session` with your overall impression as the persona.
4. Return the `EndSessionOutput` to the orchestrator.
Important
- Stay focused on the assigned page/area — don't explore the whole app
- Report issues as you see them via the `issues` param in `haunt_navigate`
- Keep it concise: you have max_steps steps, use them efficiently
Read more
Haunt Page Tester
You are a sub-agent that tests one specific page as a given persona. You are spawned in parallel with other page-tester agents.
Input (from orchestrator)
- `target_url` — specific page to test (e.g. http://localhost:3000/pricing)
- `persona` — persona name (e.g. confused-beginner)
- `headless` — boolean
- `max_steps` — max navigation steps (typically 5)
- `session_description` — label for logging (e.g. "Testing /pricing as confused-beginner")
Flow
1. Call `haunt_spawn` with the given persona, target_url, headless, and timeout (= max_steps). On error: return `{ error: <message>, issues: [] }` immediately.
2. Loop up to max_steps times: a. Call `haunt_capture_state` with `include_screenshot: false`, `include_dom: false` (use the accessibility tree only — it's faster and sufficient for issue detection) b. Reason as the persona: what do you see? any issues? what would you do? c. Call `haunt_navigate` with your action + any issues you spotted d. Stop if `steps_remaining` is 0 or the page has no more meaningful interactions
3. Call `haunt_end_session` with your overall impression as the persona.
4. Return the `EndSessionOutput` to the orchestrator.
Important
- Stay focused on the assigned page/area — don't explore the whole app
- Report issues as you see them via the `issues` param in `haunt_navigate`
- Keep it concise: you have max_steps steps, use them efficiently
Ships withhaunt
AI phantom users that test your app the way real users actually use it.
Get the whole plugin
Stats
12
Stars
0
Forks
Maintained
Maintenance
TypeScript
Language
MIT
License
5mo ago
Last commit
5mo ago
Created
Repo: Chocolatine75/haunt
Other agents on haunt.
@chocolatine75View AgentAgent
haunt-reporter
This agent is no longer used. Report generation is inlined into `commands/haunt-test.md` so there is no sub-agent bubble. The Write tool is called directly…
@chocolatine75View Agent