DAF Intent Guard for Openclaw

A sophisticated intent drift monitor for AI agents that determines if new user instructions require a minor patch, a partial rollback, or a full task abort.

deepseek609609-collab
v1.0.0
Mar 23, 2026
0
864
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install daf-intent-guard

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install daf-intent-guard using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is DAF Intent Guard?

DAF Intent Guard is a critical utility for maintaining task integrity in multi-turn AI interactions. It acts as a gatekeeper that evaluates new user input against existing task constraints and execution anchors. By calculating semantic distance and identifying the first affected checkpoint, it prevents the agent from blindly executing conflicting instructions.

This skill is part of the broader ecosystem of Openclaw Skills designed to make AI agents more robust and reliable. It provides a structured way to handle the 'change of mind' scenario, ensuring that agents only roll back to the necessary state (anchors) rather than losing all progress or continuing in an invalid direction.

DAF Intent Guard Use Cases

  • Managing mid-task modifications where a user updates search parameters or destination details.
  • Detecting conflicts between new user instructions and previously locked-in task constraints.
  • Determining the specific execution checkpoint (anchor) an agent must roll back to after a significant change.
  • Automatically aborting tasks when the new instruction fundamentally contradicts the original goal.

How DAF Intent Guard Works

  1. Before processing a new instruction in an active session, the agent calls the DAF Intent Guard with current anchors and constraints.
  2. The skill compares the old-constraints with new-constraints and analyzes the new-action.
  3. It performs a semantic distance calculation to evaluate the impact of the change.
  4. The skill returns one of three statuses: patch (update and continue), replan (rollback to a specific anchor), or abort (stop for user confirmation).
  5. The agent follows the decision to either silently update, revert its state, or prompt the user for clarification.

DAF Intent Guard Setup

To integrate this into your workflow, ensure you have Python 3 installed. You can invoke the guard via the CLI as follows:

python3 {baseDir}/daf_guard.py \
  --current-anchor <index> \
  --old-constraints '<json>' \
  --new-constraints '<json>' \
  --new-action '<action>' \
  --requires '<json_array>' \
  --anchors '<json>'

DAF Intent Guard Data Schema & Taxonomy

The skill utilizes a structured input/output model to manage task state. Key data components include:

Attribute Type Description
current-anchor Integer The index of the last successfully completed task step.
constraints JSON Key-value pairs representing the parameters of the user's request.
anchors JSON A map of previous states and their associated indices for rollback.
decision String The output status: patch, replan, or abort.

DAF Intent Guard Advanced Features

  • Semantic Distance Analysis: Sophisticated comparison of intent to prevent unnecessary task resets.
  • Anchor Rollback Logic: Identifies the exact first_affected_anchor to minimize redundant execution.
  • Seamless Integration: Designed to work within Openclaw Skills pipelines for multi-agent coordination.
  • Custom Requirement Checks: Pass a JSON array to the --requires flag to enforce specific environmental or state conditions.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Requires
Bins python3
Github Stars: 0
forks: 0

Featured*