Skills

A skill is a slash command your agent runs to drive the momentum workflow. They’re not hidden behind a CLI — they’re the primary surface the agent uses while working.

Each skill lives as a markdown file under core/commands/ (or per-agent overlay under adapters/<agent>/commands/). When the agent loads its instruction file at session start, it picks up every available skill and knows when to invoke each.

Skills group by lifecycle: Project (rare, ~once per project), Phase (every phase), Backlog (whenever), Cross-repo (only in ecosystem mode), and Quality (review, debugging). The flat list of all skills below also tells you which adapters ship each (most ship everywhere).

Project lifecycle

Skills you run rarely — typically once per project.

/brainstorm-idea

Explore any idea through structured dialogue. Nothing gets written to disk. The output is a clear, structured summary you can act on.

> /brainstorm-idea

What's the core problem or opportunity?
> need to add memory to our agent platform

Who is this for? (user, system, team)
> ...

Use this when you have a vague idea and want to think it through before committing to a project. → source

/start-project

Found the project from a clear idea — authors the charter, principles, success criteria, and roadmap from your brainstorm, then plans Phase 0. Content, not structure: momentum init owns the machinery, and foundation docs don’t exist until this command writes them (their absence is the “not founded” signal phase commands check).

> /start-project

Project name? > memory-module
Primary agent? (claude-code / codex / antigravity)
> claude-code
...

Run after /brainstorm-idea. Phase commands route here until the project is founded. → source

/migrate

Onboard an existing project with a manual or outdated momentum-like structure into proper momentum layout. Gap detection, skip-if-exists fill, index reconciliation. → source

/validate

Check the spec-structure health of a momentum project. Index-first by default; --deep for full scan.

> /validate

Reading specs/phases/index.json...
Checking active phase files... ✓
Backlog ID monotonicity... ✓
Status freshness... ⚠ Last Updated > 7 days ago

→ source

Phase lifecycle

The core loop. Run these every phase.

/brainstorm-phase

Plan the next phase through structured dialogue. Blocked by the brainstorm gate from writing to specs/ until you approve.

The brainstorm gate is a .momentum/brainstorm-active sentinel file the PreToolUse hook checks before allowing any Write/Edit on specs/ paths. Visible safety — agent can’t accidentally pre-commit to a half-finished plan. → source

/start-phase

Begin a new implementation phase. Creates the phase directory, branch, and initial commit. Then executes the plan autonomously per the contract — no per-group approval prompts, no “should I commit?” interruptions.

> /start-phase

Creating branch phase-13-site-polish...
Group 0: Mermaid tooling + baseline... ✓
Group 1: Landing rewrite... ✓
Group 2: Docs deepening... in progress
...
[Hard stop only at merge + release gate]

→ source

/sync-docs

Apply pending history entries to other specs. Reads the active phase’s history.md, identifies which other docs need updating via topics + the Affects-specs: field, applies the updates in one batch. Run at phase completion, before /complete-phase. → source

/complete-phase

Verify, finalize, and release a completed phase.

> /complete-phase

Running verification:
  npm test         → 246/246 ✓
  linkinator       → 0 broken ✓
  lighthouse       → 98 / 96 / 100 / 100 ✓

Writing retrospective.md...
Updating status.md, index.json, README.md...
Bumping package.json: 0.15.0 → 0.16.0

Ready to tag v0.16.0 and npm publish? (y/N)

The verification step is non-optional — Rule 12 (Verify Before Claim) enforces it. No release without fresh evidence captured in retrospective.md. → source

/log

Record a manual entry in the active phase history file. Use when the agent missed something the user wants tracked. Drops you into a structured prompt for the entry type, topics, and detail. → source

Backlog

/track

Track a backlog item — bug, feature, tech debt, or enhancement. One command, four item types.

> /track

Type? (bug / feature / tech-debt / enhancement)
> bug
Title? > momentum upgrade overwrites CLAUDE.md project title
Priority? (P0/P1/P2/P3) > P2
Context? > ...

Assigns the next ID in the chosen prefix, appends the row to backlog.md, notifies you of the assignment. → source

/review

Review and groom the backlog between phases. Surfaces stale items, priority drift, and missing context. Useful before /brainstorm-phase so the next phase has clean inputs. → source

Cross-repo (ecosystem mode only)

These skills only activate in ecosystem mode. In single-project mode they’re no-ops or unavailable.

/ecosystem

Cross-repo ecosystem coordination. Subcommands: init, add, remove, status, doctor. Wraps the momentum ecosystem CLI with conversational defaults. → source

/initiative

Manage cross-project initiatives. A single feature that spans more than one member repo gets one initiative file at <eco-root>/initiatives/NNNN-slug.md. → source

/session

Append a manual narrative entry to today’s ecosystem session log. Use when something happened that’s worth recording but isn’t tied to a git commit or PR. → source

/scout

Read-only context fetch from one ecosystem member repo. Returns a structured summary; writes a scout artifact for later reference. Deep dive →

/dispatch

Parallel multi-project fan-out + synthesis. One sub-agent per listed repo with auto-tailored prompts. Originating agent synthesizes the answer from the structured results. Deep dive →

/handoff

Cross-session control transfer with a structured context block. Writes <toRepo>/.momentum/inbox/handoff-NNN.md. The receiving session reads the block first. Deep dive →

/continue

Pick up a pending handoff in this repo. Reads the oldest .momentum/inbox/ handoff-NNN.md, parses the structured context, invokes the receiver flow. Idempotent. Deep dive →

Quality + debugging

/review-code

Multi-perspective code review of pending changes on the current branch. Dispatches role-specific subagents (security, QA, architecture) in parallel. Returns Critical / Important / Minor findings. → source

/systematic-debug

Systematically isolate, reproduce, and resolve task execution failures. Walks a structured debugging protocol — minimal repro, hypothesis test, fix isolated, regression check. → source

Adding your own skills

Skills are plain markdown files in core/commands/ (for cross-adapter) or adapters/<agent>/commands/ (for agent-specific). Each file has a short header explaining when to use it, then the body describing the steps the agent should follow.

The adapter contract handles where files install per agent. Authoring your own skill is on the roadmap for Phase 16 (/specify + skill authoring) — for now, write the markdown file directly and momentum upgrade won’t touch it (unmanaged files are preserved).