Skip to content

Skills

Skills are compiled, version-locked prompt artifacts stored inside your project. Running pod skill install fetches a prompt, merges in variable values you provide, renders the template, and writes the final text to skills/<slug>.md.

Skills are ideal when you want a prompt file you can commit to git, review in PRs, hand to another tool, or use without making a network call at runtime.

Why skills?

A regular pod install caches the raw prompt template — variables are still placeholders like {{name}}. A skill goes one step further: it pre-fills variables and saves the fully rendered prompt as a readable Markdown file.

pod installpod skill install
Caches the template
Persists variable values✅ in config.json
Renders to a file you can edit/commitskills/<slug>.md
Version-locked

Installing a skill

bash
pod skill install greet --name Matt

This will:

  1. Install the prompt greet (same as pod install greet).
  2. Persist { "name": "Matt" } and the pinned version to .promptodex/data/greet/config.json.
  3. Render the template and write it to skills/greet.md.

Alias: pod skill i

Options

FlagDescription
--<variable> <value>Set a template variable (persisted to config.json)
-v, --verboseShow verbose output

Skill file layout

After pod skill install greet --name Matt, your project will contain:

.promptodex/
  cache/greet/<version>.json       # raw prompt template
  data/greet/config.json           # pinned version + saved variables
skills/
  greet.md                         # fully rendered prompt

config.json looks roughly like:

json
{
  "version": "3",
  "variables": {
    "name": "Matt"
  }
}

Commit skills/ and .promptodex/data/ to git so your team uses identical rendered prompts.

Missing variables

When you install a skill:

  • Missing optional variables produce a warning and are rendered as empty strings (or their defaults, if provided).
  • Missing required variables produce an error.

You can resolve errors either by:

  1. Editing .promptodex/data/<slug>/config.json to add the missing values, then running pod skill rebuild <slug>
  2. Re-running pod skill install <slug> --var value
  3. Running pod doctor skills to see a full report across all skills

Rebuilding a skill

Re-fetch the latest version of a skill's prompt, preserve your existing variable values, and recompile the output file:

bash
pod skill rebuild greet

Use this after the prompt author publishes a new version, or after editing config.json by hand.

Version bumps

If the new version of a prompt introduces a new required variable, pod skill rebuild will warn you so you can supply the value before the skill is used.

Installing skills from a collection

Compile an entire collection of prompts as skills in one step:

bash
pod collection skill install my-code-review --author "Matt"

Every prompt in the collection is installed and rendered. Variable flags are applied to every prompt — prompts that don't declare a given variable simply ignore it. A batch report summarizes the ok / warning / error status for each skill.

Diagnosing skills

Scan every installed skill and report variable coverage:

bash
pod doctor skills

For each skill, the output is one of:

StatusMeaning
okAll required and optional variables are satisfied
warningOne or more optional variables are missing
errorOne or more required variables are missing (exits non-zero)

This is useful in CI to fail builds when a skill is missing required values.

Uninstalling

bash
pod uninstall greet

uninstall removes the prompt from promptodex.json, clears the cache, and removes any compiled skill artifacts (.promptodex/data/<slug>/ and skills/<slug>.md).

Next Steps

Released under the MIT License.