Recording risk decisions

A report is the agent's view at review time. These records are your team's decisions about it. They're kept apart from reports, so they survive the next scan and show up on the dashboards.

Which record do I need?

SituationRecordWhere
A SAST finding is real, wrong, or a risk you'll live withFinding reviewDashboard
A project doesn't meet a control, for nowExceptionsecfoo exception
You accept one specific threat from a threat modelThreat acceptancesecfoo accept
A threat was found after build that the model missedPost-build misssecfoo miss

Reviewing a SAST finding

In the dashboard, open Analysis → SAST → findings (or Ctrl+K → Review SAST findings) and pick a finding. Under the evidence, choose:

ConfirmThe finding is real and needs fixing.
False positiveThe finding is wrong.
Accept riskReal, but you'll live with it. A reason is required.

Add your name and a note. Because a finding keeps its identity across scans (it's matched by its CWE, file and the code around it, not the agent's wording), your decision stays attached when the same issue is found again. Choose Clear to send it back to awaiting review.

Exceptions: waiving a control for a project

Use an exception when a project knowingly doesn't meet a control for a while, such as during a migration. It needs a justification, an approver and an expiry date:

secfoo exception create --project . \
  --title "Legacy auth exemption" \
  --justification "Migration to SSO in progress, tracked in JIRA-123" \
  --granted-by "Jane Doe" --expires-at 2027-03-31 --control IAM

--control is free text: usually a CCM domain code (IAM, CEK…) or a standard clause such as "SOC 2 CC6.1". Add --assessment <id> to tie it to one case file. The Security Architecture dashboard shows exceptions by age (30, 60, 90 days) and highlights expired ones.

secfoo exception list --status active          # soonest-expiring first
secfoo exception update 4 --expires-at 2027-06-30 --justification "SSO cutover moved to Q2"
secfoo exception update 4 --status revoked     # the control is back in force

Threat acceptances: a named owner for one threat

When a threat model marks a threat Accepted, that's the model's judgement, not anyone's sign-off. Record who actually accepts the risk, and why:

secfoo accept create --project . --threat-id T4 \
  --title "Legacy endpoint risk" \
  --justification "Endpoint is decommissioned next quarter" \
  --accepted-by "Jane Doe" --expires-at 2027-01-31

--threat-id is the ID from the report's Threat Register. Add --run <run-id> to link the model it came from. --expires-at is a review-by date. The acceptance is shown on the Threat Modeling dashboard on every load, even after the model is rerun and its threat IDs change.

secfoo accept list --status active
secfoo accept update 2 --status revoked

Post-build misses: threats the model didn't catch

A report can't know what it missed. When an incident, pen test or later scan finds a threat that an earlier model should have caught, record it:

secfoo miss create --project . \
  --title "SSRF via webhook URL" \
  --discovered-at 2026-10-02 --discovered-by "Pen test Q3" \
  --description "Webhook target URL wasn't checked against internal IP ranges" \
  --run <run-id-of-the-threat-model>

These feed the Threats found post-build panel on the Threat Modeling dashboard (a 90-day trend and an all-time count), which is the most direct measure of how well your threat models are working.

Good practice

  • Always set an expiry. An exception or acceptance without an end date quietly becomes permanent.
  • Write justifications for a later reader: link the ticket, say what changes the decision, name the compensating control.
  • Revoke rather than delete. A revoked record keeps the history; delete removes it entirely.
  • Dates are always YYYY-MM-DD.
← Reviewing a vendor CI gates and SARIF export →