Skip to content

armgit / strix

Public Imported from GitHub
0

Open-source AI penetration testing tool to find and fix your app’s vulnerabilities.

enhance todo tool prompt

0xallam78b6c26652be
diff --git a/strix/agents/StrixAgent/system_prompt.jinja b/strix/agents/StrixAgent/system_prompt.jinja
index df5b64ac..7cf471a1 100644
--- a/strix/agents/StrixAgent/system_prompt.jinja
+++ b/strix/agents/StrixAgent/system_prompt.jinja
@@ -111,19 +111,6 @@ OPERATIONAL PRINCIPLES:
- Try multiple approaches simultaneously - don't wait for one to fail
- Continuously research payloads, bypasses, and exploitation techniques with the web_search tool; integrate findings into automated sprays and validation
-TASK TRACKING:
-- USE THE TODO TOOL EXTENSIVELY - this is critical for staying organized and focused
-- Each subagent has their own INDEPENDENT todo list - your todos are private to you
-- KEEP THE LIST SHORT-HORIZON: track only the next few concrete steps (3-6 max), not long-term goals.
-- REWRITE TODOS AS YOU LEARN: update, trim, or reprioritize the list whenever plans change or tasks finish.
-- At the START of any task: Create todos to break down your next steps into clear actions
-- BEFORE starting a task: Mark it as "in_progress" - this shows what you're actively doing
-- AFTER completing a task: Mark it as "done" immediately - don't wait
-- When you discover new tasks: Add them as todos right away and reprioritize; avoid dumping the whole project plan upfront
-- ALWAYS follow this workflow: create → in_progress → done, iterating frequently
-- A well-maintained todo list prevents going in circles, forgetting tasks, and losing focus
-- If you're unsure what to do next: Check your todo list first
-
EFFICIENCY TACTICS:
- Automate with Python scripts for complex workflows and repetitive inputs/tasks
- Batch similar operations together
diff --git a/strix/tools/todo/todo_actions_schema.xml b/strix/tools/todo/todo_actions_schema.xml
index 5cbe87b9..de32f9ee 100644
--- a/strix/tools/todo/todo_actions_schema.xml
+++ b/strix/tools/todo/todo_actions_schema.xml
@@ -1,32 +1,29 @@
<tools>
<important>
- YOU MUST USE THE TODO TOOL EXTENSIVELY — AND DO SO IN BATCHES. Minimize the number of tool calls by grouping creations and status changes together.
+ The todo tool is available for organizing complex tasks when needed. Each subagent has their own
+ separate todo list - your todos are private to you and do not interfere with other agents' todos.
- IMPORTANT: Each subagent has their own separate todo list. Your todos are private to you and
- do not interfere with other agents' todos. Use this to your advantage.
+ WHEN TO USE TODOS:
+ - Planning complex multi-step operations
+ - Tracking multiple parallel workstreams
+ - When you need to remember tasks to return to later
+ - Organizing large-scope assessments with many components
- WORKFLOW - Follow this for EVERY task:
- 1. Keep the list short-horizon: track only the next few concrete steps (3-6 max), not long-term goals.
- 2. Create/update todos as you learn new info; drop or rewrite items when plans change.
- 3. Batch operations: when setting statuses or priorities for multiple items, use ONE call with the bulk parameters (`updates` or `todo_ids`) instead of one call per todo.
- 4. BEFORE starting a task (or a small set of tasks): Mark them "in_progress" together using a single update_todo call with an `updates` array.
- 5. AFTER completing tasks: Mark them "done" together using one mark_todo_done call with `todo_ids`.
- 6. When you discover new tasks: Add them in one create_todo call (use the `todos` list) and re-prioritize; avoid giant upfront lists.
+ WHEN NOT NEEDED:
+ - Simple, straightforward tasks
+ - Linear workflows where progress is obvious
+ - Short tasks that can be completed quickly
- ALWAYS mark the current task(s) as in_progress before working on them. This shows what you're actively doing. Then mark them done when finished. Do these status updates in batches to avoid excessive tool calls.
-
- A well-maintained todo list prevents you from going in circles, forgetting important tasks,
- or losing track of your progress. USE IT CONSTANTLY.
+ If you do use todos, batch operations together to minimize tool calls.
</important>
<tool name="create_todo">
- <description>Create a new todo item to track tasks, goals, and progress. USE THIS FREQUENTLY.</description>
- <details>Use this tool liberally to create actionable items. Break down complex tasks into smaller,
- manageable todos. Each subagent maintains their own independent todo list - your todos are yours alone.
+ <description>Create a new todo item to track tasks, goals, and progress.</description>
+ <details>Use this tool when you need to track multiple tasks or plan complex operations.
+ Each subagent maintains their own independent todo list - your todos are yours alone.
- Create todos at the start of work to plan your approach, add new ones as you discover tasks,
- and mark them done as you progress. This keeps you focused, prevents you from forgetting tasks,
- and provides a clear record of what you've accomplished.</details>
+ Useful for breaking down complex tasks into smaller, manageable items when the workflow
+ is non-trivial or when you need to track progress across multiple components.</details>
<parameters>
<parameter name="title" type="string" required="false">
<description>Short, actionable title for the todo (e.g., "Test login endpoint for SQL injection")</description>
@@ -74,8 +71,8 @@
</tool>
<tool name="list_todos">
- <description>List all todos with optional filtering by status or priority. CHECK THIS OFTEN.</description>
- <details>Use this when you need fresh IDs or to reprioritize. Avoid spamming list_todos; prefer to keep a local view of the list and only refresh when state may be stale or after bulk changes.
+ <description>List all todos with optional filtering by status or priority.</description>
+ <details>Use this when you need to check your current todos, get fresh IDs, or reprioritize.
The list is sorted: done first, then in_progress, then pending. Within each status, sorted by priority (critical > high > normal > low).
Each subagent has their own independent todo list.</details>
<parameters>
@@ -107,13 +104,13 @@
</tool>
<tool name="update_todo">
- <description>Update one or multiple todo items. ALWAYS prefer bulk updates in a single call.</description>
+ <description>Update one or multiple todo items. Prefer bulk updates in a single call when updating multiple items.</description>
<parameters>
<parameter name="todo_id" type="string" required="false">
<description>ID of a single todo to update (for simple updates)</description>
</parameter>
<parameter name="updates" type="string" required="false">
- <description>Bulk update multiple todos at once. JSON array of objects with todo_id and fields to update: [{"todo_id": "abc", "status": "done"}, {"todo_id": "def", "priority": "high"}]. USE THIS INSTEAD OF multiple single-item calls.</description>
+ <description>Bulk update multiple todos at once. JSON array of objects with todo_id and fields to update: [{"todo_id": "abc", "status": "done"}, {"todo_id": "def", "priority": "high"}].</description>
</parameter>
<parameter name="title" type="string" required="false">
<description>New title (used with todo_id)</description>
@@ -146,8 +143,8 @@
</tool>
<tool name="mark_todo_done">
- <description>Mark one or multiple todos as completed in a single call. DO THIS IN BATCHES after finishing a small group of tasks.</description>
- <details>Mark todos as done right after completing them - group multiple completions into one call using todo_ids. Avoid one-call-per-item to save tool calls. This keeps your list accurate and gives you a clear picture of progress.</details>
+ <description>Mark one or multiple todos as completed in a single call.</description>
+ <details>Mark todos as done after completing them. Group multiple completions into one call using todo_ids when possible.</details>
<parameters>
<parameter name="todo_id" type="string" required="false">
<description>ID of a single todo to mark as done</description>