First successful run

A successful first run does not mean “the command exited zero.” It means your agent can prepare from your own markdown and cite the notes that matter.

Use a real project, client, research thread, or upcoming meeting. If you do not have one in mind, choose a folder or tag that has meeting notes, project notes, decisions, and open questions.

Before you start

Install Enzyme and your runtime instructions from the Quickstart. Then make sure catalyst generation has a provider path:

sign in to use Enzyme hosted credits/auth (the default for enzyme init and enzyme refresh); or
intentionally bring your own OpenAI/OpenRouter/OpenAI-compatible provider by setting env vars and passing --use-env-llm; or
configure a local OpenAI-compatible server and pass --use-env-llm if you want catalyst generation to stay local.

Default init/refresh ignore inherited LLM env keys so an existing personal API key is not used accidentally.

1. Ask for an indexability audit

Ask your agent:

Set up Enzyme for this vault. First run a read-only scan and tell me what is indexable, weakly indexable, noisy, or missing.

A useful audit should name the workspace structures Enzyme can use:

project/client folders;
meeting-note folders;
people or company pages;
tags and wikilinks;
dated notes or frontmatter;
noisy/generated folders to exclude.

If the audit cannot name real structures, do not trust retrieval yet. Fix config or add clearer markdown handles first.

2. Initialize and inspect petri

Terminal path:

enzyme scan --write-config
# review ~/.enzyme/config.toml
enzyme init
enzyme petri

The petri output should include entities you recognize. Good signs:

a project or client you have been actively working on;
a meeting folder or daily-note pattern;
a person, company, or workstream that appears in your notes;
catalysts that sound like live questions in your workspace.

Bad signs:

only generic themes;
only noisy folders;
no recognizable people/projects;
catalysts that could have been written for any vault.

3. Run the meeting-prep test

Ask your agent:

Prep me for the next meeting on <project>. Cite the notes that show unresolved commitments, decisions, or source context.

Or run directly:

enzyme catalyze "prep for next meeting on <project> unresolved commitments decisions"

A good receipt looks like this:

$ enzyme petri
entity: [[Acme onboarding]]
catalysts:
  - what remains unresolved before the next customer check-in?
  - where did pricing language change after stakeholder pushback?

$ enzyme catalyze "prep for next Acme onboarding meeting unresolved commitments"
1. projects/acme/2026-05-12 onboarding call.md
   excerpt: "Sarah asked for revised pricing language before Friday..."
   matched catalyst: what remains unresolved before the next customer check-in?

2. decisions/acme-pricing-language.md
   excerpt: "Use implementation-risk framing; avoid calling it an enterprise tax..."
   matched catalyst: where did pricing language change after stakeholder pushback?

Pass condition

The agent should answer with:

source file paths;
short excerpts;
a real unresolved commitment, decision, or project-state thread;
a sentence explaining why the files belong together;
an honest note when exact search is better for a specific name/date/title.

If it fails

Symptom	Likely cause	Fix
Petri is generic	Config selected weak/noisy entities	Edit `~/.enzyme/config.toml`, exclude noise, select better folders/tags.
No project appears	Workspace lacks handles	Add or expose project folder, wikilink, tag, or frontmatter.
Catalyze finds keyword matches only	Query too literal or catalyst set too weak	Use petri vocabulary, refresh after config repair.
Provider error during init	Missing hosted auth/credits, or incomplete env-provider config with `--use-env-llm`	Sign in for hosted credits/auth, or intentionally set provider env vars and rerun with `--use-env-llm`.
New notes do not appear	Index is stale	Run `enzyme refresh --quiet`.
Agent writes generic summaries	Writeback policy too loose	Use Capture memory as artifacts.

What not to count as success

Do not count these as success:

“Indexed N files” by itself;
an answer that cites no source files;
a list of filenames without excerpts;
generic project advice;
obvious exact matches that grep would have found faster;
invented commitments not present in notes.

The first run should earn trust by showing the user’s own notes back to them in a useful arrangement.