Add data-subject-request runx skill

crates/runx-cli/src/official_skills.rs

@@ -40,6 +40,11 @@ pub(crate) const OFFICIAL_SKILLS: &[OfficialSkillLockEntry] = &[
         version: "sha-4b5c8438cb5a",
         digest: "5af0e2dd3dd2116874e4fb886e0424f4b944c136b3430a10c31e400a0249113b",
     },
+    OfficialSkillLockEntry {
+        skill_id: "runx/data-subject-request",
+        version: "sha-6393a7059dcf",
+        digest: "23315800a48a1c6fedb1a7252ef790064ef693a0af425d2a12db3f90d4d1ed6d",
+    },
     OfficialSkillLockEntry {
         skill_id: "runx/deep-research-brief",
         version: "sha-c2d071df7f50",

dist/packets/privacy.data-subject-request.v1.schema.json

@@ -0,0 +1,47 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://schemas.runx.dev/runx/privacy/data-subject-request/v1.json",
+  "x-runx-packet-id": "runx.privacy.data_subject_request.v1",
+  "type": "object",
+  "properties": {
+    "decision": {
+      "type": "object",
+      "properties": {
+        "eligible": { "type": "boolean" },
+        "reason": { "type": "string" }
+      },
+      "required": ["eligible", "reason"],
+      "additionalProperties": true
+    },
+    "handoff": {
+      "type": "object",
+      "properties": {
+        "path": { "type": "string" },
+        "subject_id": { "type": "string" },
+        "data_classes": {
+          "type": "array",
+          "items": { "type": "string" }
+        },
+        "scopes": {
+          "type": "object",
+          "additionalProperties": true
+        }
+      },
+      "additionalProperties": true
+    },
+    "escalation": {
+      "type": "object",
+      "additionalProperties": true
+    },
+    "verdict_event": {
+      "type": "object",
+      "additionalProperties": true
+    },
+    "observations": {
+      "type": "object",
+      "additionalProperties": true
+    }
+  },
+  "required": ["decision", "escalation", "verdict_event", "observations"],
+  "additionalProperties": true
+}

packages/cli/src/official-skills.lock.json

@@ -41,6 +41,13 @@
     "catalog_visibility": "public",
     "catalog_role": "canonical"
   },
+  {
+    "skill_id": "runx/data-subject-request",
+    "version": "sha-6393a7059dcf",
+    "digest": "23315800a48a1c6fedb1a7252ef790064ef693a0af425d2a12db3f90d4d1ed6d",
+    "catalog_visibility": "public",
+    "catalog_role": "canonical"
+  },
   {
     "skill_id": "runx/deep-research-brief",
     "version": "sha-c2d071df7f50",

skills/data-subject-request/SKILL.md

@@ -0,0 +1,118 @@
+---
+name: data-subject-request
+description: Judge a data subject request, record the sealed verdict in data-store, and emit only a bounded handoff for downstream erasure or export.
+runx:
+  category: privacy
+---
+
+# Data Subject Request
+
+Judge whether a data subject request is in policy without erasing, exporting, or
+sending any data. The skill reads a request packet, requestor proof, policy
+bounds, and prior request state. It emits one typed verdict, records that verdict
+through `data-store`, and returns a bounded handoff that a separate governed
+operator run may consume.
+
+This is not a legal authority and does not replace counsel. It is an execution
+boundary for policy evidence: the receipt proves which inputs were inspected,
+which lawful basis was named, which scope was allowed or refused, and which
+durable verdict was appended.
+
+## Inputs
+
+- `request_packet`: `type`, `subject_id`, and `scope`. Scope must include the
+  request id, requested data classes, and requested scopes.
+- `requestor_proof`: `identity_provider`, `verified_at`, and `assertion`. The
+  assertion must bind the requestor to the same subject and carry a stable proof
+  reference or digest.
+- `policy`: `jurisdiction`, `trusted_identity_providers`, `lawful_bases`, and
+  `scope_bounds`. The trusted provider list is required evidence for accepting
+  the proof issuer.
+- `data_source_ref` and `store_id`: the pinned data-store binding used for the
+  subject request event stream.
+- `aggregate_id`, `expected_version`, and `idempotency_key`: the subject request
+  entity, optimistic concurrency version, and retry-safe verdict write key.
+
+## Verdict Contract
+
+The output is `runx.privacy.data_subject_request.v1` data:
+
+- `decision{eligible,reason}` always appears.
+- `handoff{path,subject_id,data_classes,scopes}` appears only when eligible.
+- `escalation` explains whether a human lane is required.
+- `verdict_event` is the exact deterministic event recorded by
+  `data-store.append_event`.
+- `observations` name the lawful-basis verdict, jurisdiction reason, verified
+  requestor reference, identity assertion digest, scope bounds, bounded handoff,
+  `aggregate_id`, `expected_version`, and `idempotency_key`.
+
+The packet must not contain an `operational_proposal` envelope. A downstream
+driver may issue a separate governed run by name, but this skill does not fire
+that rail.
+
+## State Model
+
+State is durable in `registry:runx/data-store@0.1.2`. The graph uses the
+github-sync-style shape:
+
+1. `read_projection` loads any prior decision for the subject request entity.
+2. `data-subject-request-decide` judges identity, scope, and lawful basis.
+3. `append_event` records the verdict with `expected_version` and
+   `idempotency_key`.
+4. `read_projection` reads back the durable state for receipt evidence.
+
+The aggregate id is the subject request entity, not a tenant-wide stream. The
+append is an ungated CAS write that records the decision itself; consequences
+remain downstream.
+
+## Consequence Boundary
+
+Eligible erasure returns a handoff such as
+`downstream.data-store.append_event.subject.erasure`. A downstream operator must
+issue a separate governed `data-store.append_event` run that appends a
+`subject.erasure` tombstone. Erasure is event-sourced; there is no delete
+operation in this skill.
+
+Eligible export returns a handoff to a separate read path. A downstream operator
+must run `read_projection`, `redact-pii`, and `send-as` under explicit approval.
+This skill does not read raw content, redact content, or deliver content.
+
+Ambiguous identity, missing proof, untrusted identity providers, scope disputes,
+or data classes outside `policy.scope_bounds` escalate to a human lane or block
+as `needs_agent`. The skill must refuse to invent an identity, lawful basis, or
+scope that is not grounded in the inputs.
+
+## Local Harness
+
+Run:
+
+```bash
+runx harness ./skills/data-subject-request --json
+```
+
+The inline harness has two cases:
+
+- `eligible-erasure-records-verdict`: a verified GDPR erasure request for
+  `profile` and `support_tickets` seals, emits a bounded erasure handoff, and
+  records the verdict via `data-store.append_event`.
+- `unverified-requestor-refused-no-handoff`: an unsigned email proof and an
+  out-of-scope `payment_methods` export request seal a deterministic refusal
+  naming the jurisdiction and lawful-basis/scope grounds, with no handoff.
+
+## Example Invocation
+
+```bash
+runx skill data-subject-request --json \
+  --input-json request_packet='{"type":"erasure","subject_id":"subject:customer_4242","scope":{"request_id":"dsr-2026-06-30-001","data_classes":["profile","support_tickets"],"requested_scopes":["profile","support_tickets"]}}' \
+  --input-json requestor_proof='{"identity_provider":"account-session","verified_at":"2026-06-30T14:20:00Z","assertion":{"subject_id":"subject:customer_4242","assertion_ref":"acct-session:customer_4242:2026-06-30","assertion_digest":"sha256:9d64a60e8d0c8080e88f9472c0e0ffdfad9424c2e42eb570c0f46ff3e4e73116"}}' \
+  --input-json policy='{"jurisdiction":"GDPR","trusted_identity_providers":["account-session"],"lawful_bases":{"erasure":"GDPR Article 17 erasure after withdrawn consent"},"scope_bounds":{"data_classes":["profile","support_tickets"],"erasure_allowed":["profile","support_tickets"],"export_allowed":["profile","support_tickets"]}}' \
+  -i data_source_ref=local://runx-data-subject-request/harness \
+  -i store_id=data-subject-request-harness-v1 \
+  -i aggregate_id=dsr:subject:customer_4242:request:dsr-2026-06-30-001 \
+  --input-json expected_version=0 \
+  -i idempotency_key=dsr-2026-06-30-001:verdict:v1
+```
+
+The runner computes only bounded policy checks from the supplied packet. If it
+cannot prove identity, scope, and lawful basis from those inputs, the correct
+result is a sealed refusal or human escalation, not a guessed verdict.