Paia Program Management

PAIA Program Management

Use this skill for operational project and program management across PAIA Program, PAIA Workboard, Linear, GitHub, Google Workspace, Folio, msgvault, and agent/session state.

This skill is operational. It can help inspect systems, author records, propose mutations, produce human-inspectable HTML packets, and execute approved changes.

Non-Negotiable Doctrine

No deterministic semantic logic.

Models reason. Records document and manage what the model reasons through. Scripts, hooks, and tools gather evidence, validate shape, execute approved operations, and record results. They do not classify, prioritize, route, schedule, infer health, choose approval style, or decide what matters.

If you are about to use keywords, regexes, score formulas, fixed thresholds, source precedence, or hidden category maps for meaning, stop. That is a model agency violation for this skill.

Default Workflow

1. Establish the run mode from the user's request and current context.
2. Gather evidence from the relevant systems without interpreting it in code.
3. Build a model-authored operating picture.
4. Run a research/evidence-discovery pass when the model identifies knowledge

gaps that could change the plan.

1. Deliberate through the program panel when the work is non-trivial.
2. Write the appropriate model-authored records.
3. Render static HTML inspection packets for non-trivial runs.
4. Propose concrete operations in a Change Plan before any mutation.
5. Get approval using the model-reasoned approval posture.
6. Execute exactly the approved operations.
7. Record results, failures, and repair prompts in an Execution Ledger.

Operating Modes

• triage: make scattered work inspectable.
• plan: develop initiatives, projects, dependencies, and sequencing.
• research: identify planning-relevant knowledge gaps and gather evidence

from local or external sources.

• dispatch: create/update runnable tasks and handoff packets.
• sync: reconcile drift across systems.
• review: weekly, project-level, or portfolio-level review.
• rescue: recover stale, blocked, overgrown, or poorly defined work.

Modes guide conversation and artifact expectations. They are not execution switches.

Reference Selection

• Read references/model-mediated-doctrine.md before adding scripts, hooks, or

connector behavior.

• Read references/substrates.md before touching Workboard, PAIA Program,

Linear, GitHub, Google Workspace, Folio, msgvault, or tmux/session state.

• Read references/panel-cast.md for non-trivial planning, strategy,

decomposition, research, dependency, scheduling, or rescue runs.

• Read references/artifacts.md before writing records or rendering packets.
• Read references/hooks.md before installing or changing hooks.

Templates And Scripts

Use templates in templates/ for model-authored records. Use scripts in scripts/ only for mechanical operations:

• scripts/validate-record.py: validate record shape.
• scripts/render-html-packet.py: render a model-authored record as static

HTML.

• scripts/execute-change-plan.py: refuse unapproved plans and write dry-run

execution ledgers for approved plans.

These scripts do not decide meaning.

Required Mutation Gate

Before mutating Workboard, PAIA Program, Linear, GitHub, Google Workspace, Folio, or related systems, create a Change Plan. Every operation must include:

• target system
• action name
• model rationale
• evidence references
• approval posture
• payload

Never perform these without explicit confirmation: send email, delete/archive work, close external issues, mark major work complete, cancel/move meetings, publish public material, or make broad reprioritization changes.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: dbmcco
• Source: dbmcco/claude-agent-toolkit
• License: MIT

Install and usage instructions live in the source repository linked above.