Skill

rules

View and check AI guardrails. Use when the user says /rules, asks about boundaries, or before taking potentially risky actions (deleting files, pushing code, accessing external services).

npx claudepluginhub amanasmuei/aman-claude-code --plugin aman-claude-code

Tool Access

This skill uses the workspace's default tool permissions.

Preview

You are managing the user's AI guardrails. The plugin uses the **engine v1

SKILL.md

Similar Skills

finishing-a-development-branch

174.6k

Verifies tests pass on completed feature branch, presents options to merge locally, create GitHub PR, keep as-is or discard; executes choice and cleans up worktree.

superpowers

systematic-debugging

174.6k

Guides root cause investigation for bugs, test failures, unexpected behavior, performance issues, and build failures before proposing fixes.

10 files

superpowers

writing-plans

174.6k

Writes implementation plans from specs for multi-step tasks, mapping files and breaking into TDD bite-sized steps before coding.

1 file

superpowers

Stats

Stars0

Forks0

Last CommitApr 7, 2026

Actions

View Source View Plugin View on GitHub View README

Guardrails Management

You are managing the user's AI guardrails. The plugin uses the engine v1 multi-tenant scope dev:plugin, with two possible storage locations:

Primary (engine v1, scope-aware): ~/.arules/dev/plugin/rules.md
Legacy (single-tenant fallback): ~/.arules/rules.md

Always check the primary path first. If only the legacy path exists, you can either read from it directly or recommend the user run npx @aman_asmuei/arules to migrate.

When invoked

Look for ~/.arules/dev/plugin/rules.md first; fall back to ~/.arules/rules.md.
List all rule categories and their rules.
Highlight the "Never" category prominently.

Proactive rule checking

Before taking any action that could be risky, check against the rules:

Find the rules file (primary path preferred, legacy fallback OK).
Check if the planned action matches any "Never" rules — apply keyword-overlap matching: extract meaningful words (length > 3, no stopwords) from each "Never" rule, lowercase the action, and flag if the action contains ≥2 keywords from any single rule.
If a rule matches: stop, inform the user, and ask for explicit permission.
If nothing matches: proceed normally.

If aman-mcp is registered (recommended — see the plugin README's "Live tools" section), prefer calling the rules_check MCP tool instead of doing the keyword matching yourself. The MCP tool uses the same engine library that aman-tg's production guardrails use.

Examples of actions to check:

Deleting files or data
Pushing to main/master
Modifying production systems
Accessing external APIs
Exposing secrets or credentials

If neither path exists

Proceed normally but suggest: "Consider setting up guardrails with npx @aman_asmuei/arules init to define what your AI should and shouldn't do. The new layout writes to ~/.arules/dev/plugin/rules.md (multi-tenant aware)."

Adding rules

Guide the user to:

npx @aman_asmuei/arules init — create starter rules
npx @aman_asmuei/arules add <category> — add a rule
Or edit the rules file directly (primary path preferred)