/lc-new¶
Scope a new ASTRA analysis from a research question, through
conversation. The output is a complete astra.yaml and (optionally) a
literature evidence trail. Implementation comes later — /lc-new
writes spec, not code.
Source: claude/lightcone/skills/lc-new/SKILL.md.
Allowed tools¶
Read, Write(astra.yaml), Write(universes/*), Write(CLAUDE.md),
Edit(astra.yaml), Edit(universes/*), Edit(CLAUDE.md),
Glob, Grep, Bash(astra:*), Bash(lc:*),
WebSearch, WebFetch, AskUserQuestion, Agent
Writes are locked to the spec surface — no Python, no R, no arbitrary
files. The lc-extractor subagent is dispatched via Agent.
Phases¶
- Research question. Sharpen the question, then write
version,name, anddescriptiontoastra.yamlso the user has something visible to react to from the first turn. - Analysis structure. Walk through inputs, outputs, and any
sub-analyses. One output per output: one metric, one plot, one
artifact.
inputs:andoutputs:land inastra.yamlas they crystallize. - Deep dive (per section). An optional literature pass. Collect
paper candidates with the user; for each approved paper, dispatch
one
lc-extractorsubagent in parallel. Each subagent reads the PDF, pulls verbatim quotes, runsastra paper verify-quotesto machine-verify the quotes against the source, and returns prior insights. Decisions then fall out of the conversation and the literature together. - Finalize.
astra validate astra.yaml;astra validate --verify-evidenceif quotes exist;astra universe generate -n baseline. Populate thenarrative:block (summary,methods,inputs,outputs—findingsstays TODO until results exist), then fill the## Working Notessection ofCLAUDE.mdwith conversational context the spec doesn't carry.
Writes happen at the end of each phase, not in bulk — the user always has something visible to review.
Hard restrictions (from the SKILL.md)¶
- Specification agent only. No Python, no R, no implementation code.
- Touchable files:
astra.yaml,universes/*.yaml, andCLAUDE.md(Finalize only). - Quotes are never fabricated; every evidence entry must pass
astra validate --verify-evidence. - PDFs stay inside
lc-extractorsubagents — the main agent never pulls one into its own context.
Anti-patterns called out in the prompt¶
- Bulk-writing decisions at the end instead of after each crystallizes.
- Letting vague goals like "analyze this data" pass without sharpening.
- Method-only decisions. The prompt actively probes data exclusion, variable operationalization, and inference criteria.
- Reading PDFs in the main agent context.
- Skipping
astra validate --verify-evidence.
Related¶
- After
/lc-new, ask the agent to implement the spec through the normal Claude Code workflow. /astra—astra.yamlschema, decision criteria, prior insights / findings, universe management.claude/lightcone/agents/lc-extractor.md— the literature extraction subagent definition.