Terminal
Terminal/Skill
Ranks and remediates secret incidents already flagged in your GitGuardian dashboard, through its Developer MCP server, and helps rotate exposed credentials.
By GitGuardian
Works with
npx skills add gitguardian/agent-skills --skill triage-incidentsThis skill works through the GitGuardian Developer MCP server (ggmcp) to read,
prioritize, and remediate secret incidents the GitGuardian platform has already
detected. It is distinct from scan-secrets: that finds unknown secrets in code;
this acts on known incidents in the dashboard.
This is the one MCP-first skill in the bundle. There is no ggshield subcommand for
incident triage — the work runs through ggmcp tools. ggshield is only needed for the
HMSL handoff sub-step, which is user-run regardless.
list_incidents /
get_incident. Public incidents (Public Monitoring on the worldwide perimeter:
public GitHub, gists, Docker Hub) use list_public_incidents / get_public_incident.
IDs are not interchangeable; an internal write tool called with a public ID silently
404s. Default to internal unless the user's intent is about leaks "on public GitHub /
outside the org / on Docker Hub / found by Public Monitoring".valid result raises urgency but is not a triage axis: it never lets you skip the
ownership and blast-radius questions or jump to a generic rotation plan. The deliverable
mode is selected by ownership x blast radius, and a valid production-critical credential
is Coordination (supervised, sequenced rotation), not a fast "just rotate it" — rotating
it blind can take a live system down. See doctrine principle 7.references/remediation-doctrine.md
end-to-end. Rotation-first; history-rewrite only under narrow conditions; public
exposure is always burned.get_remediation_workflow. If it returns an id, the workspace has a custom
workflow — follow it as the spine of the deliverable: render its steps verbatim and use
the doctrine to fill in the mechanics and verification under each step. If there is no
id (GitGuardian's default workflow) or the tool is unavailable, set the returned steps
aside — do not render them — and let the doctrine drive end-to-end. See doctrine § 13.remediate_secret_incidents is a read tool — not the remediation plan. Despite the
name, it changes no state: it returns occurrence data for the current repo — file paths,
line numbers, char indices — and nothing more. Ignore the remediation_instructions it
returns: do not take its remediation guidance into account. Treat this tool's output
as occurrence data only and drive the fix from the doctrine (rotation-first). Calling it
is not "remediating."unknown / no_checker /
not_checked, the follow-up is HasMySecretLeaked. Do not run ggshield hmsl yourself
and do not read the credential into context — print the command for the user
(-n none --json). If the check-hmsl skill is installed, load it for the full
protocol.scan-secrets when a scan finding turns out to already be a
tracked incident.count_incidents (or list_sources).references/gitguardian-platform.md
for auth/scope recovery and instance URLs.Follow references/triage-workflow.md — it covers the
five steps (scope → rank → drill in → drive the fix → close the loop), the
axis→filter mapping, the internal/public tool split, and scope-degradation handling.
score,
severity, source_criticality, public_exposure.incidents:write scope — re-issue the token with
write scope; see references/gitguardian-platform.md.scan-machine)..md to any page; AI-agent
index at https://docs.gitguardian.com/llms.txt).