sync-features

Sync Features Inventory

Scan the current project's codebase and update its memory stores with current feature counts and capabilities, then report what changed.

When to Use (and When NOT to)

Use this skill for	Use a simpler approach for
Syncing feature counts/inventory across a project's memory stores after real changes	Checking whether one specific feature exists — just grep/read the code
Periodic honesty pass on an inventory memory that's grown stale	A brand-new project with no memory stores yet — just write the first memory directly
Confirming a feature doesn't already exist before building it	A single quick count you won't persist anywhere

Generalized from the agent-fleet project skill. It adapts to whatever project it runs in — do NOT assume agent-fleet's base/ layout or memory names.

Step 0: Detect the project layout

Before scanning, resolve the app root and the available memory stores:

# App root: monorepo/submodule layout (e.g. agent-fleet) puts the app under base/
if [ -d base/app ]; then APP=base/app; DB=base/database; CFG=base/config; else APP=app; DB=database; CFG=config; fi
echo "APP=$APP"

• Serena memories: call mcp__plugin_serena_serena__list_memories() to see what this project actually has. Common inventory-bearing memories: architecture, implementation-status, feature_inventory, project_overview, codebase_structure — use whichever exist; do not invent ones that don't.
• Auto-memory index: the per-project MEMORY.md under ~/.claude/projects/<slug>/memory/ (loaded into context each session).
• Feature-inventory file (if the project keeps one): a *feature*inventory*.md under the project memory dir. If none exists and the project is large, propose creating one rather than assuming a path.

Step 1: Scan the codebase

Adapt these to the detected $APP/$DB/$CFG. Skip counts for things the project doesn't have (not every project has MCP tools or a versioned API):