Source: feature-gap review (upheld twice — strongest item: highest leverage, lowest risk, fully inside the state-in-repo + no-network invariants, directly serves the commercialization/promotability goal).
Problem: You capture rich append-only audit data and never analyze it. /ca:audit, /ca:status, /ca:adr-status all only list/snapshot — nothing computes a trend.
Do: add /ca:metrics (or /ca:audit --trends), read-only over the same five sources /ca:audit already cites (overrides.log, triage.log, decisions/, sprint-log.md, checkpoints/). Pin it to a small fixed set of single-number trends a lead reads in <10s, each with a direction arrow vs. the prior window:
- override-rate per N commits over the last K windows
- top-3 most-blocked gates
- small-lane share of total triage entries
- sprint low-confidence ratio
Guardrail: do NOT re-print what /ca:audit assembles — bare counts/percentages, not a second compliance packet. Answers "is discipline holding or eroding?" at a glance.
Effort: M.
Source: feature-gap review (upheld twice — strongest item: highest leverage, lowest risk, fully inside the state-in-repo + no-network invariants, directly serves the commercialization/promotability goal).
Problem: You capture rich append-only audit data and never analyze it.
/ca:audit,/ca:status,/ca:adr-statusall only list/snapshot — nothing computes a trend.Do: add
/ca:metrics(or/ca:audit --trends), read-only over the same five sources/ca:auditalready cites (overrides.log,triage.log,decisions/,sprint-log.md,checkpoints/). Pin it to a small fixed set of single-number trends a lead reads in <10s, each with a direction arrow vs. the prior window:Guardrail: do NOT re-print what
/ca:auditassembles — bare counts/percentages, not a second compliance packet. Answers "is discipline holding or eroding?" at a glance.Effort: M.