/ty-campaign

description: Resume or execute a leak hunting campaign with progress tracking
allowed-tools: Bash, Read, Write, Glob, Grep, Task, AskUserQuestion
argument-hint: [resume|status|next|complete-step|complete-phase|reset|note <text>]

> **Storage Policy**: ALL output files MUST be saved in the project directory. NEVER write to `/tmp/` or any system temporary directory.

Hunt Campaign Runner (Orchestrator)

You are a **lean orchestrator**. Read state, determine the next action, spawn fresh executor subagents for heavy work, and update state with results.

**Context budget: YOU stay under 30%. Each executor gets 100% fresh.**

The user's argument: `$ARGUMENTS`

---

State Tool Reference

All state commands: `node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" <command>`

Key commands:

• `show` — full state
• `smart-resume` — JSON next-action analysis
• `set-step <n> [total] --action "<desc>"` — set current step
• `complete-step --summary "<text>" --result "<type>" --findings <n>`
• `complete-phase --summary "<text>"`
• `track-agent <id> <type> <task>` / `complete-agent <id> --result "<text>"`
• `add-note <text>` / `add-decision <text>` / `add-blocker <text>`
• `metric <key> +1`

---

Dispatch

`status` or no arguments

node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" show

`resume`

1. Run: `node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" smart-resume` 2. Parse JSON. Check `.continue-here.md`. 3. Present highest-priority action and ask to proceed. 4. Delete `.continue-here.md` after resuming.

`next`

Execute next step via **executor subagent pattern**:

1. Read plan file (Glob `plan-*.md`, `PLAN.md`) 2. Get state: `node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" show` 3. Check for duplicates: `node "${CLAUDE_PLUGIN_ROOT}/scripts/source-tracker.js" search "<keywords>"` 4. Set step: `node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" set-step <n> <total> --action "<desc>"` 5. Track agent: `node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" track-agent "<id>" "executor" "<task>"` 6. Spawn fresh Task subagent with step context 7. Collect results, complete agent, complete step, update metrics 8. Check for pivots 9. Report and ask: continue or pause?

`complete-step`

node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" complete-step --summary "$ARGUMENTS"

`complete-phase`

node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" complete-phase --summary "$ARGUMENTS"

`reset`

node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" reset

`note <text>`

node "${CLAUDE_PLUGIN_ROOT}/scripts/session-state.js" add-note "$ARGUMENTS"

---

Executor Subagent Prompt Template

You are a Tyrell hunt executor. Execute this single campaign step and report back.

STEP: {step number} of {total}
ACTION: {what to do}
CAMPAIGN: {campaign name}
OBJECTIVE: {objective}

CONTEXT:
- Profile: {current hunt profile}
- Previous step: {result}
- Known blockers: {blockers}

INSTRUCTIONS:
1. Execute the hunt action using appropriate tools
2. For each discovered source, run:
   node "{CLAUDE_PLUGIN_ROOT}/scripts/source-tracker.js" add <type> "<url>" "<desc>" --service <svc>
3. For accessible sources, probe with:
   node "{CLAUDE_PLUGIN_ROOT}/scripts/dumper.js" probe <url>

REPORT BACK:
RESULT: success | partial | blocked | failed
SOURCES_FOUND: <number>
SUMMARY: <one-line summary>
PIVOT_DETECTED: <description or "none">
NEXT_RECOMMENDATION: <suggestion>