forgeplan decay
forgeplan decay looks ahead at the valid_until fields on every EvidencePack and
shows which decisions will lose R_eff soon. Remember: expired evidence does not
disappear — it caps at 0.1 (stale, not absent). Because R_eff is a weakest-link
function, one expiring piece of evidence can drag a previously-green decision into
the red.
This is the tool for planning a Refresh sprint: see who is about to go stale,
schedule a RefreshReport, and renew the evidence before it bites.
When to use
Section titled “When to use”- Start of the month: “which decisions will need re-verification this sprint?”
- Before relying on an old decision — is its R_eff about to drop?
- After bulk-creating evidence with the same expiry — spot the cliff before it hits.
- Session start alongside
forgeplan healthfor the full “what needs attention” view.
When NOT to use
Section titled “When NOT to use”- When you have no evidence at all (R_eff already 0) — nothing to decay.
- As a one-shot gate — use
stalefor hard expiry detection.
forgeplan decayOptions
Section titled “Options” -h, --help Print help -V, --version Print versionExamples
Section titled “Examples”See upcoming decay impact
Section titled “See upcoming decay impact”forgeplan decayOutput:
Evidence decay report─────────────────────EVID-012 expires in 14d PRD-003: R_eff 0.90 → 0.70EVID-018 expires in 30d ADR-005: R_eff 1.00 → 0.10 ⚠ cliffEVID-021 expires in 45d PRD-007: R_eff 0.80 → 0.60EVID-030 expires in 60d PRD-011: R_eff 0.90 → 0.80
4 evidence packs expiring within 60 days2 decisions will drop below 0.7 (trust threshold)Feed into a sprint plan
Section titled “Feed into a sprint plan”Combine with stale and health:
forgeplan decay # what's comingforgeplan stale # what's already expiredforgeplan health # overall blind-spot viewThose three commands are the full “refresh backlog” picture.
Output interpretation
Section titled “Output interpretation”- expires in Xd — days until
valid_until. - R_eff X → Y — current cached R_eff vs projected after decay.
- ⚠ cliff — flagged when a single expiry takes R_eff across a trust boundary (0.9 → 0.1 is a cliff; 0.9 → 0.8 is not).
- Cliffs are the priority: schedule a
forgeplan new refresh "Re-verify EVID-XXX"before the date.
How it fits the workflow
Section titled “How it fits the workflow”monthly review → decay → plan refresh sprint → renew evidence → scoreDecay is preventive. Stale is reactive. Run decay before evidence expires; run stale to catch the ones you missed.
See also
Section titled “See also”forgeplan stale— hard expiry detection after the factforgeplan score— recomputes R_eff after renewalforgeplan renew— extend valid_until on stale artifacts- Evidence methodology — decay formula and CL penalties
- CLI overview