luum-home/luum-cognitive-os/component-reality-check

Agentic Primitive Reality Check

Run the SO against itself: how many declared agentic primitives (hooks, lib, scripts, skills) actually FIRE at runtime, vs how many are declared but unused?

Not the same as the dogfood-score (composite across 7 dimensions). This skill is the drill-down into dimension 3 (hook wiring) + extends to lib, scripts, and skills directories. Use when you want the raw number, not the composite.

Today (2026-04-24 baseline): 27.7% REAL / 64.5% Dormant+Unwired.

Invocation

/component-reality-check                 # full run, default --dry-run (no persistence)
/component-reality-check --persist       # write JSONL record to .cognitive-os/metrics/aspirational-audit.jsonl for trend
/component-reality-check --strict        # exit 1 if dormant_aspirational_ratio > 0.40
/component-reality-check --worst N       # show top N worst offenders (default 10)
/component-reality-check --trend         # compare against last persisted run

What it measures

The classifier scripts/aspirational_audit.py scans every hook, lib module, script, and skill and assigns one of four labels:

Instructions

When invoked:

1. Run:

``bash uv run python3 scripts/aspirational_audit.py --dry-run --json ` (omit --dry-run if user passed --persist`).

1. Parse the JSON output and present:

• Header: total agentic primitives, measurement window
• Counts table: each label with count + percentage
• Ratio: dormant_aspirational_ratio — critical metric (target <0.40 per ADR-XX if defined)
• Worst offenders: top N (default 10) with path, type, and reason (e.g., "no JSONL output in 30d", "references missing module X")
• Recent cold-start note: any agentic primitive with mtime < 7d is expected to be DORMANT; flag those separately so they don't skew the verdict

1. If `--trend`: read previous record from .cognitive-os/metrics/aspirational-audit.jsonl and report delta (REAL +N, Dormant -N). Flag any newly-DORMANT agentic primitive.

1. If `--strict` and ratio > 0.40: exit with non-zero code so CI can gate on it.

1. Next steps section (always include): suggest actions per category:

• DORMANT >180 days: candidate for docs/99-Archive/archive/ or promotion with behavioral test.
• ASPIRATIONAL: either implement the missing dependency OR remove the reference.
• Newly added (<7d): give it time; re-check next week.

Example output

Agentic Primitive Reality Check — 2026-04-24
Measurement window: 7-30 days
 
| Label        | Count | %      |
|--------------|-------|--------|
| REAL         |   160 |  27.7% |
| DORMANT      |   305 |  52.8% |
| ASPIRATIONAL |    68 |  11.8% |
| METADATA     |    45 |   7.8% |
| Total        |   578 |        |
 
Dormant+Aspirational ratio: 0.645 (target <0.400)
Status: 🔴 ABOVE TARGET
 
Worst offenders (top 10):
  1. hooks/adr-detector.sh       DORMANT  no events 30d
  2. hooks/agent-bus-monitor.sh  DORMANT  no events 30d
  ... (8 more)
 
Cold-start (added <7d, DORMANT expected):
  - hooks/agent-quota-advisor.sh (age: 3d, ADR-056 L1)
  - hooks/agent-quota-redirect.sh (age: 3d, ADR-056 L2)
 
Delta vs last run (2026-04-17):
  REAL: +12   DORMANT: -8   ASPIRATIONAL: +3   Total: +7
 
Next steps:
  - DORMANT >180d: 47 candidates for archive/promotion (see --worst 50)
  - ASPIRATIONAL: 68 to resolve (implement deps OR remove refs)
  - Cold-start: 2 hooks acceptable, re-check 2026-05-01

Artifacts

• .cognitive-os/metrics/aspirational-audit.jsonl — trend record (append-only, --persist)
• .cognitive-os/reports/component-reality-<ts>.md — human-readable snapshot (optional, --report)

Related

• scripts/aspirational_audit.py — underlying classifier (SCOPE: os-only)
• scripts/weekly-aspirational-audit.sh — cron wrapper for automatic weekly run
• skills/dogfood-score (in progress) — composite metric; this skill is a drill-down
• ADR-031 / ADR-041 — continuous classifier design (supersedes one-shot dead-weight-audit plan)
• Engram: audit/aspirational-real — historical trend

Contextual Trigger

SO-only. Use after adding/removing agentic primitives, or on a weekly cadence via weekly-aspirational-audit.sh. Adopting projects use their own /dod-check or /readiness-check, not this skill.