pmokeDOCS
AI & Machine Resources

Agent Workflow and Versioning

Minimal-context retrieval, version pinning, and runtime verification for pmoke automation.

Bootstrap

Retrieve the concise index before loading documentation bodies:

curl -fsSL https://kerr-group.github.io/pmoke/llms.txt
curl -fsSL https://kerr-group.github.io/pmoke/ai-index.json | jq '.pmoke_version, .source_commit'

The index links directly to processed per-page Markdown. Follow only the links required for the current task; do not ingest both full-language feeds by default.

Context ladder

TaskPreferred sourceRuntime confirmation
Explain one commandPer-page CLI Markdownpmoke <command> --help
Generate configConfig JSON + JSON Schemapmoke config validate
Diagnose connectivityInstallation and troubleshooting Markdownpmoke doctor hardware
Build broad English guidancellms-en.txtRelevant native command
Compare English and Japanese termsllms-full.txtSource links and locale metadata

Version pinning

Record pmoke_version, config_schema_version, and source_commit from ai-index.json beside generated output. Reject a cached page when its SHA-256 digest does not match the corresponding manifest entry. Re-fetch the manifest when the requested pmoke version differs from the current feed.

Generation contract

Agent output must retain exact field paths, units, enum values, command spelling, and canonical source URLs. Unsupported assumptions must remain explicit. Generated configuration is a proposal until native validation succeeds.

Read the pmoke llms.txt index first. Select the smallest relevant locale-specific
Markdown source. Prefer generated JSON contracts for exact CLI and config fields.
Preserve versions, paths, units, and canonical URLs. Confirm configuration semantics
with pmoke config validate. Do not infer hardware state, credentials, or addresses.

On this page