Skip to content

Skills

Skills shape the agent's workflow around recurring research tasks. They are Claude Code and Codex slash commands. They no longer ship inside lightcone-cli. They live in the LightconeResearch/agent-skills marketplace and install as plugins.

lc init registers the marketplace in the project's .claude/settings.json. Claude Code offers to install the lightcone plugin when you trust the folder. See Install for the manual commands.

Plugins

The marketplace ships three plugins.

Plugin Skills Notes
astra astra Standalone ASTRA authoring skill. Ships the validate-on-save and activate-on-read hooks.
lightcone (recommended) new, report, cli, feedback Bundles astra. Ships the venv and session hooks plus the lc-extractor subagent. This one plugin is all you need for interactive work.
lightcone-experimental from-paper, from-code, paper-extraction, ralph, check-sentence-by-sentence, figure-comparison Opt-in. Requires lightcone. Under active development.

Skills

Plugin skills are namespaced by plugin. Install lightcone, then use these.

Skill Command What it does
new /lightcone:new Scope a research question into an astra.yaml, with optional literature extraction.
report /lightcone:report Author the MyST report that references astra.yaml elements by path.
cli /lightcone:cli The lc execution reference — spec–code invariant, status interpretation, failure diagnosis.
feedback /lightcone:feedback File a bug report with version and error context.
astra /lightcone:astra Author an astra.yaml — orientation and judgment (bundled from the astra plugin).

The lightcone-experimental plugin adds /lightcone-experimental:from-paper, /lightcone-experimental:from-code, /lightcone-experimental:paper-extraction, /lightcone-experimental:ralph, /lightcone-experimental:check-sentence-by-sentence, and /lightcone-experimental:figure-comparison.

Authoring

Skill sources, helper scripts, and authoring guidance now live in the agent-skills repository. Each skill follows the open Agent Skills standard — a SKILL.md plus optional references/, scripts/, and assets/.