SuperAPI AI Directory

lode-hard-stuff-radar

Agents & models

View repository

Updated 10-15 23:46

About this skill

Hard Stuff Radar

When code gets cheap, the valuable problems are design boundaries, risks, maintenance debt, stale assumptions, and questions that cannot be solved by generating more code. This skill surfaces those problems from Lode memory.

Inputs

  • Raw entries for the current project or selected project.
  • Artifact index entries when present.
  • Lifecycle derivation:
python <this-skill>/scripts/derive_lifecycle.py --vault <vault> --slug <project-slug>

Output Contract

Group findings into:

  • Recurring open questions
  • Architecture or contract risks
  • Abandoned alternatives worth revisiting
  • Maintenance debt signals
  • Problems needing tests, design, research, or human judgment

Every finding must cite raw entry timestamps or artifact ids. If evidence is weak, label it as inferred.

Rules

  • Raw entries are the fact source.
  • Artifact index entries are supporting source navigation.
  • Do not invent hard problems from artifact titles alone.
  • Missing artifact index must not block output.
  • Prefer a short, actionable radar over a broad generic risk list.