Claude Code Slash Commands Template Directory

Template Directory

Start with seven reusable commands that map to common development workflows. Save them as markdown files under .claude/commands/ or convert them to .claude/skills/<name>/SKILL.md as the workflow grows.

• code-review.md: inspect the current diff and return prioritized findings.
• prd.md: turn a feature idea into a compact product requirements draft.
• refactor.md: propose and apply a scoped simplification plan.
• test.md: find, run, and fix targeted tests.
• security.md: review auth, injection, secrets, and data exposure risks.
• changelog.md: summarize user-visible changes since the last release.
• commit-message.md: draft a conventional commit message from the current diff.

Code Review Template

Use this for /code-review-local when you want a lightweight project command rather than the bundled /code-review skill.

---
description: Review the current git diff for correctness issues
allowed-tools: Read, Grep, Glob, Bash(git diff *)
---

Review the current changes for bugs, regressions, missing tests, security concerns, and unnecessary complexity.

Context:
- Changed files: !`git diff --name-only HEAD`
- Diff: !`git diff HEAD`

Return findings first, ordered by severity. Include file paths and exact lines when possible.

PRD Template

Use this for /prd when product ideas need to become implementation-ready requirements.

---
description: Draft a concise PRD from a feature idea
argument-hint: [feature or problem]
allowed-tools: Read, Grep, Glob
---

Create a PRD for: $ARGUMENTS

Include:
1. Problem and target user
2. Goals and non-goals
3. User stories
4. UX states and edge cases
5. Technical notes from the repo
6. Acceptance criteria
7. Open questions

Refactor And Test Templates

Use focused templates for refactor and test workflows so Claude Code does not drift into unrelated cleanup.

# .claude/commands/refactor.md
---
description: Plan and apply a scoped refactor
argument-hint: [file or area]
allowed-tools: Read, Grep, Glob, Edit, Bash(npm test *)
---

Refactor $ARGUMENTS without changing behavior. First explain the smallest safe plan, then edit, then run the most relevant test.

# .claude/commands/test.md
---
description: Run and repair targeted tests
argument-hint: [test pattern or area]
allowed-tools: Read, Grep, Glob, Edit, Bash(npm test *)
---

Find tests related to $ARGUMENTS, run the narrowest useful command, fix failures caused by the current changes, and summarize what remains.

Security, Changelog, And Commit Templates

These templates are copy-ready and intentionally conservative about tools.

# .claude/commands/security.md
---
description: Review changes for security risks
allowed-tools: Read, Grep, Glob, Bash(git diff *)
---

Review the diff for auth bypass, injection, exposed secrets, unsafe deserialization, SSRF, path traversal, and data leakage.

# .claude/commands/changelog.md
---
description: Draft changelog notes from the current diff
allowed-tools: Bash(git log *), Bash(git diff *)
---

Draft concise changelog bullets grouped by Added, Changed, Fixed, and Migration Notes.

# .claude/commands/commit-message.md
---
description: Draft a conventional commit message
allowed-tools: Bash(git diff --cached *), Bash(git status *)
---

Write one conventional commit subject and a short body based on the staged diff.

Official Claude Code References

The official SDK slash commands page documents markdown command files, frontmatter, arguments, bash execution, file references, and namespacing. The skills guide explains why skills are now the recommended format for richer custom commands.

Common Errors

Keep command files boring and explicit. The more a command mutates state, the more carefully its tool permissions and review steps should be documented.

• Using the same filename for different workflows and confusing the slash command name.
• Adding broad Bash permission when a narrow git or test command is enough.
• Depending on unstaged changes when the command reads only staged diff.
• Forgetting to run /reload-skills or restart the session after adding files.

FAQ