| name | projects-setup |
| description | This skill should be used when the user wants to start a new project-portfolio
for partner project-portfolio steering — cogni-projects models consultants,
projects, and staffing, so new portfolio work starts here. Trigger on: "set up
a projects portfolio", "start a project portfolio", "new cogni-projects
portfolio", "initialize project-portfolio steering", "create a staffing
portfolio", or any request to begin structured consultant/project/staffing
work — even if the user does not say "setup" explicitly.
|
| allowed-tools | Read, Write, Edit, Bash, Glob, Grep |
Projects Setup
Initialize a new cogni-projects portfolio: a self-contained directory holding
consultants, projects, and assignments that later skills (staffing match,
backfilling recommender, partner-meeting dashboard) read and write. This skill
is the foundation entry point — run it once per portfolio before any staffing
logic exists.
Core concept
A portfolio is one cogni-projects/<portfolio-slug>/ directory rooted by a
projects-portfolio.json manifest. The manifest is the source of truth for the
portfolio's identity and holds the (initially empty) consultants[],
projects[], and assignments[] entity lists that downstream skills populate.
A .metadata/ directory carries the append-only logs those skills write to.
Setup is idempotent: re-running against an already-initialized portfolio
returns {"success": false, ...} with an "already initialized" message and
never overwrites existing data.
Workflow
Step 1: Gather portfolio identity
Determine two values from the user's request (ask only for what is missing):
- Portfolio name — the human-readable label (e.g. "Nordic Advisory 2026").
- Portfolio slug — a kebab-case identifier derived from the name (e.g.
nordic-advisory-2026). Derive it automatically from the name unless the user
supplies one.
Step 2: Confirm before writing
Echo the resolved slug and name back to the user and confirm the target
directory cogni-projects/<slug>/ before running the init script.
Step 3: Scaffold the portfolio directory
Run the init script (resolved against CLAUDE_PLUGIN_ROOT, with a plugin-cache
fallback so it works whether the plugin is loaded locally or from cache):
bash "${CLAUDE_PLUGIN_ROOT:-$(ls -td "$HOME"/.claude/plugins/cache/*/cogni-projects/*/ 2>/dev/null | head -1)}/scripts/portfolio-init.sh" "<portfolio-slug>" "<portfolio-name>"
The script returns a single JSON line {"success": bool, "data": {...}, "error": "string"}:
success: true — the portfolio was scaffolded. data.path is the new
directory; data.slug echoes the slug.success: false with an "already initialized" error — the portfolio already
exists; nothing was changed. This is a clean no-op, not a failure.
Step 4: Summarize and point to next steps
Report the created directory and outline what comes next: define the data model
and author entities, then run the staffing match engine. Keep the summary short.
Output
cogni-projects/<portfolio-slug>/
├── projects-portfolio.json Root manifest (identity + consultants/projects/assignments)
├── consultants/ Consultant entity records (populated later)
├── projects/ Project entity records (populated later)
├── assignments/ Assignment records (populated later)
└── .metadata/ Append-only logs (execution, staffing, decisions)
