Prefix Commands
NepWalk uses prefix commands, never slash commands. A prefix is plain text you type into the chat; any agent (Claude Code, Codex, OpenCode, Gemini, Cursor) reads it from AGENTS.md and acts.
See
nepwalk-docs/AGENTS.md §0for the project-wide rule.
Why prefixes, not slash commands
Prefix (idea:) |
Slash (/idea) |
|
|---|---|---|
| Works in Claude Code | ✓ | ✓ |
| Works in Codex / OpenCode / Gemini | ✓ | ✗ |
| Single source of truth | ✓ (AGENTS.md) |
✗ (one file per tool) |
| Stays in sync across tools | ✓ | ✗ |
| Team can switch agents freely | ✓ | ✗ |
If a tool offers a slash-command feature — ignore it. Type the prefix.
Inline context
Every prefix accepts free-text after the colon. The text is seed input for the interview, not the final answer.
idea: trip management for Nepal tour operators
prd: also consider offline mode and version log
srs: nestjs + fastify + prisma + postgres
tasks: backend auth phase
acceptance: backend
The interview still asks about every template section the inline context does not cover.
Full prefix list
See AGENTS.md §6 in nepwalk-management/ for the authoritative table. Highlights:
| Prefix | Purpose |
|---|---|
idea: [context] |
Start idea intake interview |
idea: build |
Write the doc from completed answers |
idea: promote v<N> |
Bump status: draft → review |
prd: [context] / prd_new: |
Generate or version-bump PRD |
srs: [context] / srs_new: |
Generate or version-bump SRS |
user-story: / roadmap: |
Generate from latest PRD/SRS |
products: [context] |
Fan-out: PRD + SRS + user stories + roadmap |
marketing: [context] |
Fan-out: ICP + positioning + messaging + GTM |
tasks: [context] |
Task division across departments |
tasks: <dept> [context] |
Task division for one department |
acceptance: [context] |
Acceptance criteria across departments |
acceptance: <dept> [context] |
Acceptance criteria for one department |
deps: [context] |
Dependency map |
management_status: |
Health check |
management: tasks |
List unclaimed tasks |
management: init-docs |
Create all required output docs from templates |
management: start <task> |
Begin a task (worktree, branch, context load) |
management: agent <slug> <task> |
Spawn specialist sub-agent |
management: publish <file> |
Promote draft to review + open PR |
state: |
Regenerate PROJECT_STATE.md |
What : means
The colon is a signal to the agent: "this is a structured command, not a question." Anything after it is context for that command. A line without a known prefix is treated as plain conversation.