Projects and Priorities
Every project has a history. Every status update builds on a previous one. Every risk assessment references a prior decision. And every time a professional starts a project briefing from scratch — reconstructing context from memory, re-explaining background to a new audience, rediscovering a risk that was already known — they are experiencing project amnesia.
Industry surveys suggest that senior professionals spend two to four hours per week on project context reconstruction: finding out where a project stands, what was last decided, what is currently at risk. Layer 3 of work.local.md holds this context persistently — so the briefing never starts from zero.
This lesson builds Layer 3. After Lesson 3 (Layers 1 and 4) and Lesson 4 (Layer 2), this is the final layer. When you finish here, your work.local.md foundation is complete.
What Layer 3 Contains
Layer 3 is the project memory layer. For each active project, it stores the working intelligence that would otherwise exist only in your head: the codename, the current milestone, what is at risk, and the key decisions already made.
The official productivity plugin's task management handles tasks — what needs to be done. Layer 3 handles projects — what is in flight, what matters, and what is at risk. These are different levels of the same picture.
When /agentic-office:task-intelligence classifies a task as P1 because "it belongs to a P1 project," it is reading from Layer 3. When the daily digest reports that Project Nighthawk is at risk, it is reading from Layer 3. When a search for "Karachi expansion" surfaces project context alongside meeting notes and terminology, part of that comes from Layer 3.
The quality of the layer determines the quality of everything that reads from it.
The Project Entry Format
Each project entry has ten fields. Here is the format from the workplace-context skill:
- name: "[Project full name]"
codename: "[Internal codename if different]"
status: "[PLANNING / IN PROGRESS / AT RISK / BLOCKED / COMPLETE]"
priority: "[P1 / P2 / P3]"
owner: "[Named person]"
description: >
[What this project is; why it matters; key context]
current_milestone: "[What is happening now]"
next_milestone: "[What comes next; by when]"
at_risk: "[What could go wrong; what is currently stalled]"
decisions:
- "[Decision made; approximate date]"
key_contacts:
- "[Person; role in project]"
Two fields deserve special attention.
at_risk: This is the most important field and the most commonly underspecified. "Some scheduling challenges" is not useful. "Facility agreement negotiations stalled — 10+ days with no update, blocking Q3 target" is useful. A risk you do not write down is a risk you will forget. Write it specifically.
decisions: Key decisions already made. Not the full discussion — just the decision, in one line. "Decided to proceed with Chapter 39 before Chapter 38" or "Facility agreement: escalated to COO on 14 March." These entries prevent decisions from being relitigated when context changes. They also surface in search results, making it possible to answer "when did we decide X?" without digging through meeting notes.
Building the Case Study Entries
Here are the three project entries from the Panaversity case study.
AgentFactory — AI Agent Factory Book (P1)
- name: "AI Agent Factory"
codename: "AgentFactory"
status: "IN PROGRESS"
priority: "P1"
owner: "Zia Khan"
description: >
Open-source curriculum for building and monetising AI agents.
Book: agentfactory.panaversity.org. Current focus: completing
Part 3 domain chapters. Target: Q2 2026 curriculum launch.
current_milestone: "Chapter 39 (Productivity & Agentic Office) — in progress"
next_milestone: "Chapter 40 (Complete AI-Native Organisation) — due May 2026"
at_risk: "Chapter sequencing — skills files must accompany each chapter;
delay in skills build delays chapter publication"
decisions:
- "Chapters include both narrative and downloadable skills library"
- "Each chapter ends with exercises of 60-90 minutes each"
- "Skills files follow router + products + agents + template format"
key_contacts:
- "Dr. Sana Mirza — Head of Curriculum (PHM framework ownership)"
- "Omar Farooq — analytics and student performance data"
Project Nighthawk — Karachi Expansion (P2)
- name: "Karachi Expansion"
codename: "Project Nighthawk"
status: "AT RISK"
priority: "P2"
owner: "COO, PIAIC"
description: >
Expansion of PIAIC operations to a new Karachi campus.
Target: Q3 2026 launch. Key dependency: facility agreement.
current_milestone: "Facility agreement negotiation — STALLED"
next_milestone: "Signed facility agreement — required before Q3 target"
at_risk: "Facility agreement negotiations stalled — 10+ days with no update.
Escalation needed; Q3 target at risk if not resolved this week."
decisions:
- "Q3 2026 launch target confirmed at Executive Weekly (February 2026)"
- "Escalation to COO triggered 14 March 2026 — no response yet"
key_contacts:
- "Local government liaison — facility agreement"
- "Facilities team — site preparation"
BankersAI — Banker AI Workshops (P2)
- name: "Banker AI Workshops"
codename: "BankersAI"
status: "RECURRING"
priority: "P2"
owner: "Zia Khan"
description: >
Monthly AI upskilling workshops for banking sector professionals.
Uses Cowork as the primary platform demonstration.
Builds Panaversity's presence in the financial services sector.
current_milestone: "Workshop #7 — topic: Digital FTEs for compliance teams"
next_milestone: "Workshop #7 delivery — last Saturday of March 2026"
at_risk: "Content review must happen 7 days before delivery — confirm date"
decisions:
- "Workshop content reviewed internally 7 days before each delivery"
- "Cowork designated as primary platform for all workshop demonstrations"
key_contacts:
- "Workshop team — content preparation and logistics"
Priority Levels
Every project carries a P1, P2, or P3 classification. The levels follow the same logic as task priorities — but applied to projects rather than individual tasks.
P1 — Highest priority. Critical to the organisation's strategic goals. A significant problem at a P1 project has real consequences: missed revenue, reputational damage, a strategic commitment at risk. AgentFactory is P1 because it is the organisation's primary product and revenue driver.
P2 — Important, not existential. Significant to the organisation's goals but not immediately critical. A slip causes a problem; it does not cause a crisis. Project Nighthawk and BankersAI are P2. Important — but a problem at either would not put the organisation at risk.
P3 — Standard. Valuable but deferrable. Can slip without immediate consequence.
If you classify more than three projects as P1, you have not prioritised — you have listed. The rule is simple: if everything is critical, nothing is. Forced-choice is the discipline. Classifying a fourth project as P1 should require explicitly demoting one of the existing three. If you cannot make that trade-off, reconsider whether the fourth project truly meets the P1 criteria.
Codenames and Why They Matter
Project Nighthawk is not a code. It is a professional practice.
Codenames serve three purposes:
Privacy. "Project Nighthawk" is safe to say in a meeting where not everyone needs to know about the Karachi expansion. "We are evaluating Karachi" is not. This is not secrecy — it is appropriate information control. Not every attendee at the monthly workshop needs to know about the facility negotiations.
Efficiency. A short, memorable name is faster to use and harder to mishear or mistype. "AgentFactory" is faster than "the AI Agent Factory book project" in every context where it is understood.
Identity. A codename creates a shared reference point. When the team refers to "Nighthawk" rather than "the Karachi project," the codename carries all its associated context. It becomes a unit of communication.
In Layer 3, the codename field ensures the search layer recognises both the full name and the codename. Searching for "Nighthawk" and searching for "Karachi expansion" should surface the same entry.
Introducing Workplace Search
With all four layers now built or building — Layer 1 (personal) from Lesson 3, Layer 4 (organisational) from Lesson 3, Layer 2 (team) from Lesson 4, and Layer 3 (projects) from this lesson — a new capability becomes available: cross-context search.
/agentic-office:workplace-search searches all four layers simultaneously. It does not search one layer and return results — it searches all of them and groups results by layer, giving you the complete picture of what your workplace memory knows about any topic.
Here is a search against the case study data. Note that this query uses "Islamabad expansion" — a topic referenced in the organisational layer and meeting notes, but not yet a project entry:
/agentic-office:workplace-search
> What do we know about the Islamabad expansion? Search everything.
Sample search output:
SEARCH RESULTS: "Islamabad expansion"
════════════════════════════════════════════════════════════
Found in:
PROJECT MEMORY:
BankersAI — Islamabad is a potential third workshop location (after
Karachi and Faisalabad); no project record created; classified as
under consideration
MEETING NOTES / DECISION LOG:
Executive Weekly [March 2026]: Islamabad expansion deferred pending
Nighthawk resolution. Decision D-003. No timeline set.
D-003 [March 2026]: DEFERRED — Islamabad workshop expansion
Trigger for revisit: Project Nighthawk facility resolved
PEOPLE MEMORY:
No person entries with primary ownership of Islamabad expansion
TERMINOLOGY:
PIAIC Faisalabad SIG: AI Special Interest Group for business leaders
in Faisalabad — a similar structure could be the model for an
Islamabad SIG if expansion proceeds
TASK / DELEGATION LOG:
No active tasks related to Islamabad expansion
RECOMMENDED NEXT STEP:
Create a project stub for Islamabad if you want to track it formally.
Currently a deferred decision — nothing will happen until Project
Nighthawk is resolved. Add to P3 with status DEFERRED.
════════════════════════════════════════════════════════════
This output assembled information from three different layers — project memory, meeting notes (organisational layer), and terminology (organisational layer) — and synthesised them into a complete picture. It also flagged what is missing: no project record, no named owner, no active tasks. The search output is not just a retrieval; it is an assessment of completeness.
The recommended next step is the search layer's most useful feature: it does not just tell you what you know — it tells you what to do next.
What the Search Cannot Find
The search only surfaces what is in work.local.md. When the output says "not in workplace memory," it is accurate — and it is useful. It means one of three things:
- The information does not exist (the event has not happened yet)
- The information exists but was not recorded (gap to fill)
- The information is stored elsewhere (Notion, email, Slack) and needs to be added
The search makes your memory gaps visible. Every "not in workplace memory" result is an invitation to add the missing information.
Building Your Own Project Entries
Now build Layer 3 for your own context. Aim for three to five active projects — the initiatives you are currently responsible for or closely involved with.
For each project, two disciplines:
Be honest about the at_risk field. The most common mistake is writing a risk that is too vague to act on. "Some dependencies" is not a risk entry. "External vendor contract renewal due 1 April — no conversation started yet" is a risk entry. If you cannot name the specific risk, you probably have not thought it through. Write it anyway — the act of writing reveals what needs to be clarified.
Log at least one decision per project. The discipline of decision-logging is about preventing rework. If you have already decided something — about scope, approach, timeline, or method — write it in the decisions field. Future-you will spend time relitigating it if you do not.
Exercise: Building Layer 3 and Testing Search
Time: 35 minutes
Plugin commands: /agentic-office:workplace-context, /agentic-office:workplace-search
Step 1 — Add 3–5 project entries (20 minutes)
For each active project, use:
/agentic-office:workplace-context
> Add project: [Project name]
Codename: [Internal codename]
Status: [PLANNING / IN PROGRESS / AT RISK / BLOCKED / COMPLETE]
Priority: [P1 / P2 / P3]
Owner: [Named person]
Description: [What this is and why it matters]
Current milestone: [What is happening now]
Next milestone: [What comes next; by when]
At risk: [What is stalled or could go wrong]
Decisions: [Key decisions already made]
After adding all entries, verify your P1 count. If you have more than three P1 projects, force a reclassification.
Step 2 — Test workplace search (15 minutes)
Run three search queries against your work.local.md:
/agentic-office:workplace-search
> What do we know about [your most important project]?
/agentic-office:workplace-search
> What is at risk across all my projects right now?
/agentic-office:workplace-search
> Find everything related to [a person from Layer 2] across all projects.
For each query, evaluate: is the result complete? What is missing? What would you need to add to work.local.md to make future searches on this topic more comprehensive?
Deliverable: Layer 3 of your work.local.md with 3–5 project entries including honest at_risk fields and at least one decision per project. At least one search query evaluated for completeness with identified gaps.
After Lesson 3 (Layers 1 + 4), Lesson 4 (Layer 2), and this lesson (Layer 3), your work.local.md foundation is complete. All four layers are in place. Every subsequent lesson in this chapter builds on this foundation — task intelligence, delegation, digests, meetings, and the executive dashboard all read from the memory you have built here.
Try With AI
Use these prompts in Cowork or your preferred AI assistant.
Reproduce: Add the AgentFactory project entry and test the search.
Add the following project to my workplace memory:
Name: AI Agent Factory
Codename: AgentFactory
Status: IN PROGRESS
Priority: P1
Owner: Zia Khan
Description: Open-source curriculum for building and monetising AI agents.
Current focus: Part 3 domain chapters. Target: Q2 2026 launch.
Current milestone: Chapter 39 (Productivity & Agentic Office) — in progress
Next milestone: Chapter 40 (Complete AI-Native Organisation) — due May 2026
At risk: Chapter sequencing — skills files must accompany each chapter
Decisions:
- Chapters include narrative content and downloadable skills library
- Each chapter ends with exercises of 60-90 minutes each
Now search: What is the status of AgentFactory?
Verify the output includes status, current milestone, and the at_risk field.
What you are learning: The search output should surface the project entry fields directly — including the at_risk field. If the at_risk field does not appear in the search results, either it was not saved correctly or the search query needs to be more specific. This exercise tests the round-trip: add → verify → search → evaluate.
Adapt: Add your own P1 project and test completeness.
Add my most important current project to workplace memory:
[Describe your P1 project — name, what it is, current milestone,
what is at risk, at least one key decision already made]
After adding it, search: What do we know about [project name]?
Evaluate: Is anything missing that you expected the search to surface?
What additional context would make future searches on this project
more complete?
What you are learning: The gap between what you expected the search to find and what it actually returned is a direct measure of your memory completeness. Every gap is a prompt to add something to work.local.md. This iterative process — search, notice the gap, add — is how workplace memory improves over time.
Apply: Run a risk assessment across all projects.
Search my workplace memory: What is at risk across all my projects
right now? Surface every at_risk field, every stalled item,
every decision that is pending.
After reviewing the results:
1. Which risk has the highest urgency and why?
2. Which risk has been unaddressed the longest?
3. What would you recommend as the first action on the highest-priority risk?
Then tell me: what information is NOT in workplace memory that would
make this risk assessment more complete?
What you are learning: A cross-project risk assessment is one of the most valuable outputs of Layer 3. The question "what is NOT in workplace memory" trains you to use the search layer not just to retrieve information but to identify what you have not yet recorded. The most important risks are often the ones that have not been written down yet.
Flashcards Study Aid
Continue to Lesson 6: Task Intelligence →