| name | implementation-planning |
| description | Use when producing a scoped implementation plan from approved priorities. Creates file-level specificity with parallelization analysis. |
Implementation Planning
Overview
Take approved priorities from Discovery and produce a concrete, scoped implementation plan that specifies exactly what changes to make, to which files, in what order, and what can be parallelized.
When to Use
- In the Planning phase, after the product owner approves Discovery priorities
- When the product owner asks "how would we implement X?"
Process
- Read approved priorities: Load
workspace/pas-development/{slug}/discovery/priorities.md
- For each priority, determine scope:
- Which existing files need modification? (Read them first)
- What new files need creation? For new PAS artifacts (skills, agents, processes), specify the
pas-create-* command with all flags. Scripts live at
plugins/pas/processes/pas/agents/orchestrator/skills/creating-{type}/scripts/pas-create-{type}.
Use --base-dir to target the correct root. If the creation scripts themselves are the change
target, note this as a bootstrap exception and create manually.
- What files might be affected indirectly? (Check cross-references)
- Identify dependencies: Which changes depend on others? Which are independent?
- Assign to agents: Based on the change type:
- Architectural changes (process definitions, orchestration, library skills) → Framework Architect
- Documentation, tutorials, naming, readability → DX Specialist
- Signal processing, changelog updates → Feedback Analyst
- PRs, issue comments, linking → Community Manager
- Determine parallelism: Group independent changes that can be dispatched simultaneously
- Estimate scope: Flag any priority that seems too large for a single cycle — suggest splitting
Output Format
# Implementation Plan
## Priorities Addressed
{List from Discovery, with reference}
## Changes
### Change 1: {description}
**Priority:** {which Discovery priority this addresses}
**Agent:** {who implements this}
**Files:**
- Modify: `{path}` — {what changes}
- Create: `{path}` — {what this is}
Command: `bash plugins/pas/.../pas-create-{type} --name ... --base-dir ...` (for PAS artifacts)
**Depends on:** {other change numbers, or "none"}
### Change 2: ...
## Execution Order
### Parallel Group 1 (no dependencies)
- Change 1 → Framework Architect
- Change 3 → DX Specialist
### Parallel Group 2 (depends on Group 1)
- Change 2 → Framework Architect
### Sequential
- Change 4 → Feedback Analyst (after all other changes)
## Out of Scope
{Anything considered but deferred, with reasoning}
Quality Checks
- Every priority from Discovery is addressed (or explicitly deferred with reasoning)
- Every file path is verified to exist (for modifications) or verified not to exist (for creations)
- No circular dependencies between changes
- Agent assignments match agent capabilities
- Plan is achievable in a single Execution phase — if not, split into sub-cycles
Common Mistakes
- Creating PAS artifacts by hand instead of using
pas-create-* scripts. The scripts guarantee
correct structure (SKILL.md frontmatter, changelog.md, feedback/backlog/.gitkeep). Manual
creation risks missing convention requirements.
- Library skills:
pas-create-skill outputs to processes/{p}/agents/{a}/skills/{name}/ — there
is no --library flag. For library skills, create under a temporary agent path then move to
library/, or scaffold the directory manually following the same structure.