Problem
The Leadtype skill is easy to activate for prompts that name Leadtype features such as MDX conversion, llms.txt, linting, or docs-site integration. It is less reliable for ordinary authoring prompts such as:
- “add a troubleshooting page”
- “document this new option”
- “rewrite the quickstart”
- “review these docs for agent readability”
Those tasks still need Leadtype-specific guidance when the repository depends on leadtype, contains docs.config.ts / leadtype.config.ts, or uses Leadtype MDX components and frontmatter. The current .agents/skills/leadtype/SKILL.md frontmatter does not explicitly include writing, editing, reviewing, or maintaining documentation.
This is an activation problem, not a distribution problem. #15 covered capability distribution and discovery channels; the skill description and authoring workflow still need to make the correct routing decision obvious.
Blume’s core skill is a useful reference: its description explicitly names “writing Markdown/MDX content” alongside setup and configuration.
Proposal
Expand the trigger description
Make the description explicitly cover:
- writing, editing, reviewing, or restructuring Markdown/MDX docs in a Leadtype project;
- editing frontmatter, navigation, includes, type tables, and Leadtype MDX components;
- maintaining docs after product/API changes;
- running the relevant lint, generation, and agent-readability checks.
The repository context should be part of the trigger. A prompt does not need to say “Leadtype” when the project has a Leadtype dependency or config file.
A possible direction:
Work with Leadtype-powered documentation: write, edit, review, and validate Markdown/MDX; use Leadtype components and frontmatter; configure navigation and sources; generate agent-readable artifacts; and integrate or debug a Leadtype docs pipeline. Use for documentation work whenever the project depends on leadtype or contains docs.config.ts / leadtype.config.ts, even if the user does not name Leadtype.
Add a docs-authoring route to the skill body
Add an explicit “Writing or editing docs” path that tells the agent to:
- Detect the installed Leadtype version and project config.
- Read the smallest relevant bundled topic, including authoring/frontmatter/component guidance.
- Follow the project’s existing MDX conventions.
- Run config-aware linting and generation after changes.
- Inspect generated markdown/agent artifacts when the edit affects agent readability.
Keep progressive disclosure: the skill should route to the relevant bundled page rather than loading the whole docs bundle.
Align every Leadtype-owned discovery surface
Review and align:
.agents/skills/leadtype/SKILL.md in this repository;- the auto-generated docs skill in
packages/leadtype/src/llm/skills.ts;
- the root
AGENTS.md block emitted by leadtype init;
- packaged/bundled skill snapshots and docs.
The existing AGENTS.md pointer already says “using leadtype or writing/editing docs”; the skills should make the same promise.
Add activation evals
Add a small routing eval set with positive and negative prompts.
Positive examples in a Leadtype project:
- add a guide;
- document a config option;
- reorganize navigation;
- fix broken docs links;
- improve an MDX page for agents;
- update docs after a release.
Negative examples:
- generic prose editing outside a Leadtype project;
- documentation work in a repository with no Leadtype signal.
Acceptance criteria
Problem
The Leadtype skill is easy to activate for prompts that name Leadtype features such as MDX conversion,
llms.txt, linting, or docs-site integration. It is less reliable for ordinary authoring prompts such as:Those tasks still need Leadtype-specific guidance when the repository depends on
leadtype, containsdocs.config.ts/leadtype.config.ts, or uses Leadtype MDX components and frontmatter. The current.agents/skills/leadtype/SKILL.mdfrontmatter does not explicitly include writing, editing, reviewing, or maintaining documentation.This is an activation problem, not a distribution problem. #15 covered capability distribution and discovery channels; the skill description and authoring workflow still need to make the correct routing decision obvious.
Blume’s core skill is a useful reference: its description explicitly names “writing Markdown/MDX content” alongside setup and configuration.
Proposal
Expand the trigger description
Make the description explicitly cover:
The repository context should be part of the trigger. A prompt does not need to say “Leadtype” when the project has a Leadtype dependency or config file.
A possible direction:
Add a docs-authoring route to the skill body
Add an explicit “Writing or editing docs” path that tells the agent to:
Keep progressive disclosure: the skill should route to the relevant bundled page rather than loading the whole docs bundle.
Align every Leadtype-owned discovery surface
Review and align:
.agents/skills/leadtype/SKILL.mdin this repository;packages/leadtype/src/llm/skills.ts;AGENTS.mdblock emitted byleadtype init;The existing
AGENTS.mdpointer already says “using leadtype or writing/editing docs”; the skills should make the same promise.Add activation evals
Add a small routing eval set with positive and negative prompts.
Positive examples in a Leadtype project:
Negative examples:
Acceptance criteria
leadtypedependency/config) are documented as sufficient activation context.leadtype initpointer, generated skills, and repository skill use consistent language.