Skip to content
Forgeplan

forgeplan_decay

Show how aged / expired evidence has degraded R_eff scores across the workspace. For each affected decision, it reports the current R_eff (with decay penalties applied) vs a hypothetical fresh R_eff (if the expired evidence were renewed). This quantifies the “silent quality loss” caused by evidence aging.

Category: Quality

When an agent calls it

Section titled “When an agent calls it”
  • Quality audit — see which decisions are silently losing trust from old evidence.
  • Prioritising refresh work — the biggest current-vs-fresh deltas are the highest-impact renewals.
  • Before a release — make sure shipped decisions aren’t riding on decayed evidence.
  • Explaining R_eff drops to the user (“why did PRD-039 score go from 0.85 to 0.41?”).

Decay follows the R_eff rule: expired evidence is not removed — it drops to CL0 with weight 0.1, so the weakest link in the chain collapses the overall score.

Input parameters

Section titled “Input parameters”

No input parameters. Call this tool with an empty object {}.

Returns

Section titled “Returns”
{
  "decayed": [
    {
      "id": "PRD-039",
      "kind": "prd",
      "r_eff_current": 0.41,
      "r_eff_fresh": 0.85,
      "delta": -0.44,
      "expired_evidence": [
        { "id": "EVID-047", "days_expired": 12 }
      ]
    }
  ],
  "total_impacted": 1
}

Example invocation

Section titled “Example invocation”
{}

Typical sequence

Section titled “Typical sequence”
  1. forgeplan_stale — find expired artifacts.
  2. forgeplan_decay — see which decisions they damaged.
  3. Sort by delta descending — highest-impact first.
  4. forgeplan_renew or create new EvidencePack → forgeplan_link.
  5. forgeplan_score — confirm R_eff rebounded.

CLI equivalent

Section titled “CLI equivalent”
Terminal window
forgeplan decay

See also

Section titled “See also”