Skip to content

SKILL.md

Latest commit

 

History

History
49 lines (38 loc) · 2.16 KB

SKILL.md

File metadata and controls

49 lines (38 loc) · 2.16 KB
Raw
name labnb-idea
description Record a promising but not-yet-implemented experiment idea in the global lab notebook index.

Capture Idea

This is a flat, standalone labnb-idea skill so every AI coding agent can discover it directly. It is a focused companion to the broader labnb skill.

Helper scripts referenced as skills/labnb/scripts/... ship with the installed labnb skill; resolve them relative to that skill's directory (for example ~/.claude/skills/labnb/scripts/ or ~/.agents/skills/labnb/scripts/), not the repository layout.

Use this skill when there is a useful direction to remember, but not enough evidence, time, or scope to run it now.

Guardrails

  1. Respect any parent constitution, project policy, or task-level write constraint already in scope.
  2. Review the notebook index first so the idea does not duplicate an existing entry unnecessarily.
  3. Capture speculative directions as ideas instead of opening active experiments that would encourage unsafe or premature writes.
  4. Record idea status explicitly, defaulting to ideation unless a better status is known.
  5. Treat provenance as best-effort; deletions still require explicit confirmation.
  6. When writing or referring to provenance, use W3C PROV-O terms.
  7. Review and update the idea's local rules and durable memory before any substantive follow-up action.

Flow

  1. Review the parent constitution and local project guardrails.
  2. Summarize existing entries for the project first.
  3. Choose a concise slug for the idea.
  4. Register the idea:
python skills/labnb/scripts/register_experiment.py \
  --lab-root "$LAB_ROOT" \
  --project-slug "$PROJECT_SLUG" \
  --experiment-slug "$IDEA_SLUG" \
  --objective "$OBJECTIVE" \
  --entry-kind idea \
  --overall-budget "$OVERALL_BUDGET" \
  --loop-budget "$LOOP_BUDGET"
  1. Fill in idea.md with:
    • why it matters
    • what prior evidence to revisit
    • the smallest useful first slice
    • what would justify promoting it into an experiment
  2. Fill in memory.md with durable facts, pending waits, and the safest resume point.

Ideas should be first-class index entries, not buried in an active experiment log.