Gates tools (Pro)
Risk classes and approval gates. This group has 6 tools, all of them available only with spec-editor Pro.
| Tool | Edition | Summary |
|---|---|---|
gate_auto_decide |
Pro | Ask the project’s LLM to decide a pending gate request on the owner’s behalf (the UI’s ‘Auto’ button next to Approve/Reject). |
gate_decide |
Pro | Approve or reject a pending gate request. |
gate_status |
Pro | Status of a gate (or all gates seen) for one element, folded from gate/decision journal events, plus this element’s decisions log (gate_type, verdict, by,… |
gates_pending |
Pro | List pending gate approval requests (review/plan/release). |
plan_gates_reevaluate |
Pro | Re-evaluate pending plan gates that were escalated only because no root source was found (empty root_source_ids): a class A/B gate whose roots now all resolve… |
risk_assess |
Pro | Compute the deterministic risk class (A-D) for an element: a source (blast radius from its violates leaves) or a plain requirement/work-item element (blast… |
gate_auto_decide
Pro tool. Available with the spec-editor Pro plugin.
Ask the project’s LLM to decide a pending gate request on the owner’s behalf (the UI’s ‘Auto’ button next to Approve/Reject). Gathers the request’s facts (element, origin(s), related elements, possible duplicates, deterministic risk), asks the model for a JSON verdict (approve/reject/unsure + confidence + rationale), and always journals a ‘gate/auto_review’ audit event. WRITE tool, human-only: only applies the verdict (as ‘human:auto-review’) when apply=true, the verdict isn’t ‘unsure’, and its confidence is high enough — otherwise the request is left pending and only the audit trail records what the model would have said.
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
apply |
boolean | no | Apply the verdict if confident enough (default true); false always leaves the request pending — the audit event is still journaled either way. |
element_id |
string | yes | Element id |
gate |
string | yes | Gate name (review/plan/release/…) |
project_path |
string | yes | Path to spec-editor project (directory with methodology.yaml) |
gate_decide
Pro tool. Available with the spec-editor Pro plugin.
Approve or reject a pending gate request. WRITE tool, human-only for human gates (agents/policies are rejected with an error).
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
by |
string | no | Approver identity, e.g. ‘human:vscode’ (default ‘human:mcp’) |
decision |
string | yes | ‘approved’ or ‘rejected’ |
element_id |
string | yes | Element id |
gate |
string | yes | Gate name (review/plan/release/…) |
project_path |
string | yes | Path to spec-editor project (directory with methodology.yaml) |
gate_status
Pro tool. Available with the spec-editor Pro plugin.
Status of a gate (or all gates seen) for one element, folded from gate/decision journal events, plus this element’s decisions log (gate_type, verdict, by, author_role/verifier_role, rationale, evidence_refs, rounds) — who decided each gate and on what evidence. Read-only.
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
element_id |
string | yes | Element id |
gate |
string | no | Gate name (review/plan/release/…); omit for all |
project_path |
string | yes | Path to spec-editor project (directory with methodology.yaml) |
gates_pending
Pro tool. Available with the spec-editor Pro plugin.
List pending gate approval requests (review/plan/release). Applies any newly-answered gate questions first.
Parameters
This tool takes no parameters.
plan_gates_reevaluate
Pro tool. Available with the spec-editor Pro plugin.
Re-evaluate pending plan gates that were escalated only because no root source was found (empty root_source_ids): a class A/B gate whose roots now all resolve owner_approved is approved by policy; any other stays pending with the found roots recorded. Gates with recorded roots are untouched. Every re-evaluation is journaled as ‘plan/gate-reevaluated’. WRITE tool.
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
by |
string | no | Operator identity (default ‘human:mcp’) |
project_path |
string | yes | Path to spec-editor project (directory with methodology.yaml) |
risk_assess
Pro tool. Available with the spec-editor Pro plugin.
Compute the deterministic risk class (A-D) for an element: a source (blast radius from its violates leaves) or a plain requirement/work-item element (blast radius from its own implements/consists_of/children, one level deep). Pure function of the model — no LLM call, nothing written.
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
element_id |
string | yes | Element id |
project_path |
string | yes | Path to spec-editor project (directory with methodology.yaml) |