Pages
SLA Rubric Config v1

Source: c79 finding-lifecycle — the rubric as org config · 2026-08-25

The frameworks are agnostic — every risk program must have an SLA rubric, but each org defines its own targets (Mark M, 2026-08-20). So the rubric is the one genuinely new config surface the lifecycle needs. Edit a target and the consequence panel re-judges the live world: SLA data = f(findings, rubric, period), with the rubric as the knob.

Program presets:
Remediation targets
critical
high
medium
lowadvisory

c78 placeholder 7 / 30 / 90 / 180 — superseded by Mark M's actual program (2026-08-20 call): 14 / 30 / 90 / advisory.

Consequences · live world re-judged90% within SLA · 10 judged · 0 past target
F-1041highsecrets.yaml read in agent session2d / 30d · 27d left
F-1027mediumSandbox disabled on harness20d / 90d · 69d left
F-1038mediumMCP server added without review — pg-tools4d / 90d · 85d left
F-1039mediumPrompt file contains internal hostname patterns4d / 90d · 85d left
F-1022 · accepted risk — clock paused, expires 2026-10-01
Clock policy · fixed rulings, not configuration

Clock starts at detection · exception pauses with mandatory expiry · expiry re-opens mechanically · reclassify re-baselines from original detection · accepted risk never counts as resolved. These are the semantics of the machine, ruled on the anatomy page — making them configurable is the first step toward the workflow tooling this cycle deliberately doesn't build.

Composition notes

  • The consequence panel is the argument — editing a number visibly moves breach counts and the within-SLA verdict because deriveWorld takes the rubric as an argument. Nothing is stored; the page is the function.
  • Presets demonstrate program variance — Mark M's own numbers, the superseded c78 placeholder, and the 48-hour-criticals program he described ("HIPAA HITRUST… criticals in 48 hours, no negotiate"). Same findings, three different judgments.
  • "Advisory" is a first-class target valuedays: null, per Mark M's low severity ("when we can get to it"). Advisory findings age informationally and can never breach.
  • Changing a rubric after a quarterly export silently rewrites submitted history — target days become a contract at first submission. Product needs rubric versioning eventually; out of mock scope, recorded as an open thread.

Qpoint Brand Style Guide