Install This Asana Skill Set for AI
- 6 hours ago
- 8 min read
Lately, I’ve been experimenting with automating parts of my workflow using Codex and Claude Cowork, especially in ways that connect to Asana (PS if you want to know more about Codex, here's my article about when I installed it a few weeks ago).
What interests me most is not automation for the sake of automation. It’s using AI to reduce the kind of work that pulls me out of strategic thinking: checking for updates, organizing tasks, chasing follow-up, sorting priorities, and manually building visibility across work that is already happening.
I’ve been testing an Asana skill set built by Eli Bosley, Director of Product at Lime Technology Inc., that allows these tools to interact with Asana in plain English. The result is not “AI doing my job for me.” It is AI helping me see what matters faster, follow up more consistently, and protect more time for deep work.
What I’m using it for right now
1. A Friday follow-up review Each Friday, I want a summary of all the tasks I’m part of that are not assigned to me but may still need my follow-up. This helps me catch loose ends without manually combing through every project and thread.
2. Importing important tasks into Asana from my personal email If something important lands in my personal email, I want a path to turn it into an Asana task so it lives inside that system instead of me needing to remember to look there.
3. A monthly People + Compliance review I’m using automation to send myself a monthly review that includes:
new compliance items that may apply to my workplace
people-related tasks that are stalled
issues that may need attention before they become bigger problems
For HR and People leaders, this kind of review can be incredibly helpful. It creates a recurring checkpoint instead of relying on memory. I created my last article with information that was shared with me when I tested out this automation.
4. Structuring My Tasks for better decision-making I want my Asana My Tasks list organized in a way that helps me work more intentionally. I’m using tags to show me:
Quick Wins
Waiting on Others
Needs Clarity
Delegate to Human
Delegate to AI
Deep Work
This is one of my favorite use cases because it shifts Asana from being just a list of work into a tool that helps me decide how to work. My goal is simple: spend as much time as possible on deep work and reduce the friction around everything else.
5. Weekly staff updates I’m also experimenting with a weekly update that gives me visibility into my staff:
what they completed this week
what is stalled
what is coming up next
This kind of summary is useful because it creates a lightweight management dashboard without requiring constant manual reporting.
Why this matters
I think AI Automation can feel like a huge undertaking. What I'm finding as I start to experiment is that the most useful applications I’m finding are more simple but attached to outcomes that are better than what I can do as a human. AI can see into the systems where I participate in a way that I just can't. The value is not just that something gets done faster; it is that I am reducing the cognitive load of managing work.
Instead of:
checking multiple places for status
trying to remember who is waiting on what
organizing a bunch of check-in/reminder subtasks
manually reorganizing my task list
letting important but non-urgent work disappear
…I am creating systems that surface the right information at the right time.
Like what you're reading and want to stay in touch? Subcribe to the newsletter here.
How to set up automations in Codex and Claude Cowork
At a high level, the process is:
0. Decide whether you want to install the skill set Before anything else, review what the skill set does and decide whether you want to give your AI tool access to Asana in this way. The GitHub documentation explains the capabilities in detail, but the decision is ultimately yours. As with any integration, it is worth understanding what access you are granting and whether it makes sense for your workflow, your comfort level, and your organization’s practices.
1. Connect the tool to Asana The first step is creating an Asana personal access token and connecting it to the skill so the agent can access your tasks, projects, and workflows.
2. Install the skill into your AI environment From there, the skill can be installed into Codex, Claude Code, or both, depending on how you work.
3. Use plain English requests Once installed, you can prompt it in natural language to do things like:
show you your tasks
organize or tag your task list
pull project summaries
build briefings
import tasks from structured data
generate workflow-specific summaries
4. Build recurring automations around your real workflow This is where it becomes powerful. The real opportunity is not just asking for one-off help. It is setting up recurring patterns:
weekly reviews
monthly compliance check-ins
follow-up summaries
task sorting rules
manager updates
5. Test the automation This part really matters. Don’t assume the first version is right just because it runs. Test whether it is actually pulling the right tasks, using the right criteria, and producing something useful. Review the output, look for what is missing, and notice where it is over-including or under-including information. Give it direction on how to improve if it needs it. Watch how it explains what it is doing. Some of the things that are problems for it might not make sense to you (that happens to me a lot). You can learn more about the issue that is coming up by asking your AI, and it is possible to ask it to troubleshoot the issue and resolve it even if you don't fully understand what the issue is.
6. Ask the AI to improve the automation over time One of the most useful parts of this process is that you can refine it collaboratively. Ask the AI to improve the prompt, tighten the logic, reduce noise, and learn from what was or was not helpful in previous outputs. The best automations are usually not built in one pass. They get better through iteration.
In other words, don’t just use AI as an assistant. Use it to create systems.
Best practices for AI prompts
The quality of the output depends a lot on the quality of the prompt. I've added my People + Compliance Monthly Update prompt at the end of this artice as an example. A few best practices I’ve found helpful:
Be very clear about the outcome Don’t just say “review my tasks.” Say what you want the review to produce.
For example:
identify stalled items
show what needs my follow-up
group tasks by action needed
flag anything unclear or blocked
Define the audience and the purpose Is this summary for you? For a manager? For a team meeting? For a compliance review? The tool performs better when it knows the lens it should use.
Give categories and rules If you want tasks sorted into buckets like Quick Wins or Deep Work, define what those mean. Don’t assume the model will interpret them the way you do.
Tell it what to exclude This is a big one. If you only want tasks you’m collaborating on but that are not assigned to you, say that. If you don’t want completed tasks, say that. If you only want stalled items older than a certain number of days, include that.
Use recurring prompts for recurring decisions If you find yourself making the same judgment every week, that is often a sign that the decision can be systematized.
Use AI to make the prompt better. After you've written out what you want it to do, ask AI to improve the prompt.
My bigger takeaway
What I’m learning is that a way to effectively start building AI workflows is with things that are practical, repetitive, operational, and tied to an outcome that you desire (not just an output). They help you:
see what needs attention
reduce time spent triaging work
create consistency
protect time for the work only you can do
aggregate information quickly
I've shared my People + Compliance prompt below, but if any of the other ones I created are of interest to you, I’m happy to share them. Send me a DM and let me know what interests you.
Many thanks to GitHub, Asana, Anthropic, and ChatGPT for the tools that make this possible, and to Harvard University for their course on Agentic AI that has helped me learn so much in so little time.
Example: People + Compliance Prompt
Create a task in my Asana assigned to me on the first business day of each month called Monthly People + Compliance Review.
Use the automation memory first so the run learns from prior approvals and does not repeat stale items.
Create the main Asana task first. In the task description, include a structured summary of updates, trends, and action items that are relevant to my work as a VP of People & Culture and HR strategist. Focus on practical, decision-useful information that may affect how I do my job, support leaders, stay compliant, or shape people strategy.
For external compliance and employment-law coverage:
* only include items that apply to companies with 25 employees or fewer
* disregard federal-contractor-specific items unless I explicitly ask for them
* do not repeat the same compliance item month after month just because it still exists
* include a repeated item only when there is a material update, a new law or deadline, or I likely need a reminder because the action window is still open or approaching
* when a repeated item is included for reminder purposes, treat it like a reminder/update, not brand-new news
Include:
* Employment law or compliance updates that may affect California employers, remote teams, multi-state employers, or international contractors/employees, subject to the small-employer and no-federal-contractor filters above
* Notable HR, people, leadership, compensation, benefits, payroll, or workplace policy trends
* Relevant developments in AI, workplace technology, or automation that may affect HR, operations, or how organizations manage people
* Best practices or emerging thinking related to performance management, compensation, hiring, leadership development, employee relations, culture, and organizational design
* Anything especially relevant for small to midsize companies, remote companies, or growing tech organizations
Also review only relevant Asana work, with emphasis on:
* tasks assigned to me
* tasks where I am a collaborator
* tasks in people, hiring, operations, leadership, compensation, performance, or planning-related projects
* tasks or discussions that mention specific employees, contractors, role changes, compensation, performance, staffing, leaves, growth, succession, organizational changes, or other people-related decisions
In the main task body:
* include clickable links to original external sources whenever possible
* keep the formatting compact and easy to scan, without extra blank lines between topics
add a short section called *Suggested next steps for compliance items** with concise, practical next steps for each compliance item that is surfaced
add a short section called *What this may mean for me** with 3-5 practical implications
add a short section called *Possible action items** with concrete follow-up ideas if applicable
For internal people topics:
* create concise subtasks under the main monthly review task instead of only listing them in the description
* each subtask should cover one high-signal employee, contractor, hiring, promotion, staffing, compensation, planning, policy, or org-design topic that likely needs my attention
* prioritize signal over detail
* keep summaries concise
* include current status where possible
* include why it matters
* link directly to the relevant Asana task or tasks
* avoid unnecessary sensitive detail
* surface only what I likely need to review or act on
* treat the monthly review subtasks as review markers, not completion markers for the underlying people item
* if a prior monthly review subtask was checked off but the original source Asana task is still open, continue surfacing that item again when it is still reminder-worthy or materially updated
* only treat an employee topic as complete and safe to suppress when the underlying source Asana task is complete or the latest source context clearly shows the issue is resolved
Keep the update concise, useful, and easy to scan. Prioritize high-impact information over generic news.
