Agent-Skills.md

Agent Skills: mantic-sh

Use Mantic.sh for context-aware code search in local repositories. Use when an agent needs intent-based file discovery, symbol definition/reference lookup, impact analysis, or session-aware context retrieval before coding.

UncategorizedID: xiaoland/agents/mantic-sh

Repository

xiaoland/agents
xiaoland

Install this agent skill to your local

pnpm dlx add-skill https://github.com/xiaoland/agents/tree/HEAD/skills/mantic-sh

Skill Files

Browse the full folder contents for mantic-sh.

Download Skill

Loading file tree…

skills/mantic-sh/SKILL.md

Skill Metadata

Name
mantic-sh
Description
Use Mantic.sh for context-aware code search in local repositories. Use when an agent needs intent-based file discovery, symbol definition/reference lookup, impact analysis, or session-aware context retrieval before coding.

mantic-sh

Simple guidance for deciding when to use Mantic.sh and how to run it effectively.

When to use this skill

  • Find code by intent when exact keywords are unknown.
  • Locate likely files before reading a large codebase.
  • Jump to symbol definitions across packages.
  • Find symbol references for refactors or code review.
  • Estimate blast radius with impact-aware search.
  • Keep retrieval context across a multi-step coding task.

Common trigger phrases:

  • "where is X implemented"
  • "find auth/payment/router logic"
  • "go to definition of X"
  • "find references of X"
  • "show impact of changing X"

When not to use Mantic.sh

  • Exact literal or regex text search (prefer rg).
  • Mass text operations (for example, scanning all TODO comments).
  • Interactive fuzzy browsing (prefer fzf style workflows).

Decision flow

  1. If the task is exact text or regex matching, use rg first.
  2. If the task is intent-based discovery, use mantic <query>.
  3. If a symbol name is known, use mantic goto <symbol> or mantic references <symbol>.
  4. If change risk matters, add --impact.
  5. If results are weak, retry with better domain terms or --semantic.

Procedure

  1. Run Mantic without installation:
npx mantic.sh@latest "<query>" --files
  1. Choose output mode:
  • --files for fast path lists
  • --json for structured processing
  • --markdown for readable terminal output
  1. Use symbol intelligence when applicable:
npx mantic.sh@latest goto "<SymbolName>"
npx mantic.sh@latest references "<SymbolName>"
  1. Use deeper analysis only when needed:
npx mantic.sh@latest "<query>" --impact
npx mantic.sh@latest "<query>" --semantic
  1. Restrict scope for precision:
  • --path <dir>
  • --code, --test, --config
  1. Open top-ranked files and continue with normal edit/test workflow.

Completion checks

  • Query type matches the selected search mode.
  • At least one relevant file or symbol location is found.
  • Impact was checked before risky changes.
  • Fallback to rg was used if Mantic output is empty or too broad.

Quick examples

npx mantic.sh@latest "stripe payment webhook" --files
npx mantic.sh@latest goto "UserService"
npx mantic.sh@latest references "handleLogin"
npx mantic.sh@latest "session auth middleware" --impact --json

MCP note

If your environment already exposes Mantic MCP tools (search_files, get_definition, find_references), prefer those tools over shell commands in that environment.