Skip to content

/mako-trace

TRIGGER when: user asks to trace a specific route, schema object, file, table, RPC, or error string through its evidence. Covers `route_trace`, `schema_usage`, `file_health`, `auth_path`, `trace_file`, `preflight_table`, `trace_edge`, `trace_error`, `trace_table`, `trace_rpc`.

shell
$ npx -y skills add drhalto/agentmako --skill mako-trace --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/mako-trace
How auto-invocation works

Context preview

The summary Claude sees to decide when to auto-load this skill.

TRIGGER when: user asks to trace a specific route, schema object, file, table, RPC, or error string through its evidence. Covers `route_trace`, `schema_usage`, `file_health`, `auth_path`, `trace_file`, `preflight_table`, `trace_edge`, `trace_error`, `trace_table`, `trace_rpc`.

SKILL.md

mako-trace.SKILL.md
description: >-
  TRIGGER when: user asks to trace a specific route, schema object, file, table,
  RPC, or error string through its evidence. Covers `route_trace`,
  `schema_usage`, `file_health`, `auth_path`, `trace_file`, `preflight_table`,
  `trace_edge`, `trace_error`, `trace_table`, `trace_rpc`.
when_to_use: >-
  Use after discovery has identified a concrete entity or when the user already
  provided a route, file, schema object, table, RPC, edge, auth surface, or
  error text.
allowed-tools: "mcp__mako-ai__*"

Mako Trace

Use this skill for focused evidence traces around a known target. The output should help the user understand what implements the target, what depends on it, and what evidence supports the answer.

Targeted Answer Tools

`route_trace`

Use to find the route handler, nearby files, and evidence for a route/API behavior.

  • Best when the user names a route, page, method/path pair, or route-like URL.
  • Pair with `auth_path` for authorization questions.
  • Pair with `route_context` when the user wants a wider route neighborhood.

`schema_usage`

Use to find where an indexed schema object, table shape, or validation type is defined and referenced.

  • Best for schema objects and type-ish entities.
  • Use before editing validation contracts or generated schema surfaces.
  • Pair with `trace_table` or `table_neighborhood` for database-backed entities.

`file_health`

Use to understand a file's role, dependents, risks, and likely blast radius.

  • Best before editing a central or unfamiliar file.
  • Pair with `imports_impact` for a deeper dependency view.
  • Do not use it for repo-wide orientation; use `repo_map`.

`auth_path`

Use to inspect likely authentication or authorization boundaries for a route, file, feature, or flow.

  • Best when the user asks what protects something or whether auth is enforced.
  • Pair with `route_trace`, `route_context`, and `tenant_leak_audit` when the

risk is tenant or data-boundary related.

  • If no exact route, file, or feature matches, `auth_path` returns

`matched: false`, `reason`, and a suggested `cross_search` fallback instead of throwing a batch-breaking error.

  • Do not overclaim; report evidence and uncertainty.

Composer Trace Tools

`trace_file`

Use for a compact file trace: symbols, imports, routes, schema touches, and notable relationships before editing.

`preflight_table`

Use before changing database-backed behavior to inspect table usage, RLS, relations, and common query paths.

`trace_edge`

Use to gather evidence for a relationship between two entities, files, routes, tables, symbols, or RPCs.

`trace_error`

Use to investigate an error string, stack, or failure mode and produce likely causes with evidence.

`trace_table`

Use to trace a table through schema, code references, routes, and related RPCs.

`trace_rpc`

Use to trace a database RPC/function through schema and app-code callers.

Feedback Logging

Log `agent_feedback` when a trace here was notably useful, partial, noisy, stale, wrong, or wasted the turn. Skip routine calls.

Required procedure (see `/mako-ai:mako-guide` for full rules and reason-code vocabulary):

1. Call `recall_tool_runs` to get the prior run's `requestId`. Do not fabricate one — if no run is recalled, skip feedback. 2. Call `agent_feedback` with `referencedToolName`, `referencedRequestId`, `grade: "full" | "partial" | "no"`, `reasonCodes` from the starter vocabulary in `/mako-ai:mako-guide`, and a short `reason`.

See Also

  • Use `/mako-ai:mako-neighborhoods` when a table, route, or RPC needs wider

bundled context.

  • Use `/mako-ai:mako-graph` when the question is dependency path, flow, impact,

or change planning.

  • Use `/mako-ai:mako-workflow` when the user wants a review bundle,

verification bundle, or pre-ship artifact after tracing.

Read more
Read it on GitHub ↗
Ships withagentmako

agentmako is a local-first codebase intelligence engine for AI coding tools. It gives agents like Codex, Claude Code, Cursor, and local harnesses a compact Reef-first tool surface for understanding a project before they edit it.

Get the whole plugin, auto-invoked
Stats
51
Stars
0
Views
5
Forks
Active
Maintenance
TypeScript
Language
Apache-2.0
License
7d ago
Last commit
3mo ago
Created

Repo: drhalto/agentmako

Other skills on agentmako.