forgeplan blindspots

Find blind spots — active decisions that have no evidence backing them (R_eff = 0) and orphan artifacts not linked to any parent. This is the most important health-triage command: an active PRD/RFC/ADR without evidence is a “false promise” — a decision that looks real but has no measurement behind it.

When to use

Session start — run this right after forgeplan health and fix before starting new work (do not accumulate debt)
CI gate — fail the build if any active critical artifact has no evidence
Pre-release — zero blind spots before cutting a tag
Retro — count how many blind spots accumulated over the sprint

Not to use when

You want stuck items (different problem) → use forgeplan blocked
You want overall rollup → use forgeplan health
You want decision timeline → use forgeplan journal with --risk

Usage

forgeplan blindspots

Options

  -h, --help     Print help
  -V, --version  Print version

Examples

Print all blind spots:

forgeplan blindspots

Use as a CI gate (exits non-zero on any blind spot):

forgeplan blindspots || exit 1

Combined with health --ci for full gating:

forgeplan health --ci && forgeplan blindspots

Output interpretation

Two sections — blind decisions and orphans:

BLIND DECISIONS (active, R_eff = 0.00)
  PRD-046  Docs v0.18.0 catch-up    → create EvidencePack + link
  ADR-008  Cloudflare Pages choice  → create EvidencePack + link

ORPHANS (no parent)
  PRD-044  Unused PRD               → link to Epic or deprecate
  NOTE-031 One-off observation      → link to artifact or ignore

Summary: 2 blind decisions, 2 orphans

Section	What it means
Blind decisions	Active PRD/RFC/ADR with `R_eff = 0` (no evidence)
Orphans	Artifact with no `parent` and no inbound link

Exit code: 0 if clean, 1 otherwise. Use this in git pre-push hooks or CI.

How it fits

blindspots is a strict subset of what health reports — extracted as a standalone command because it is the most common gate to enforce. The unified workflow mandates:

session start → health → blindspots → (fix) → new work

If blindspots is not empty, fix it first. Never let active decisions drift without measurement — that is exactly what the R_eff weakest-link model is designed to prevent.