blitzstrike 2.4.174 → 2.4.176

package/CHANGELOG.md

@@ -5,6 +5,25 @@ All notable changes to Blitz Strike are documented here.
 The format follows [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/).
 
+## [2.4.176] — 2026-10-11
+
+### Added — retest context (completes the retest loop)
+
+- New `retest_context` tool: given a confirmed/still_vulnerable/inconclusive
+  finding, return everything needed to re-verify it — finding key fields, PoC
+  evidence grouped by role (proof/baseline/verification) + reproducible flag,
+  the active operation constraints, and the current goal state.
+
+## [2.4.175] — 2026-10-11
+
+### Added — evidence roles (structured proof/baseline/verification/supporting)
+
+- Evidence records now carry a `role` (proof = trigger request/response, baseline =
+  normal request, verification = negative control/marker, supporting = anything else).
+- `finding_attach_evidence` accepts a `role`; new `finding_evidence_roles` tool
+  returns the role breakdown + a `reproducible` flag (true only when a finding has
+  proof + baseline + verification — the three roles a HackerOne-grade PoC needs).
+
 ## [2.4.174] — 2026-10-11
 
 ### Added — observation ledger + lineage (third write-back channel)

package/dist/index.js

@@ -33211,6 +33211,7 @@ function makeEvidence(input) {
   return {
     evidence_id: nextEvidenceId(),
     type: input.type,
+    role: input.role ?? "supporting",
     description: redactSecrets(input.description),
     source: input.source ? redactObject(input.source) : undefined,
     sink: input.sink ? redactObject(input.sink) : undefined,
@@ -33236,6 +33237,18 @@ function verifyEvidence(evidence) {
   }
   return { ok: failed.length === 0, failed };
 }
+function evidenceRoles(evidence) {
+  const roles = { proof: 0, baseline: 0, verification: 0, supporting: 0 };
+  for (const e of evidence)
+    roles[e.role] = (roles[e.role] ?? 0) + 1;
+  return {
+    roles,
+    proof: roles.proof > 0,
+    baseline: roles.baseline > 0,
+    verification: roles.verification > 0,
+    reproducible: roles.proof > 0 && roles.baseline > 0 && roles.verification > 0
+  };
+}
 
 // src/invariants.ts
 function checkFindingInvariants(finding) {
@@ -66758,23 +66771,87 @@ ${statusLine}` : statusLine },
   });
   server.registerTool("finding_attach_evidence", {
     title: "Attach evidence to a finding",
-    description: "FINDINGS: attach a redacted, SHA-256-tagged evidence record to an existing canonical finding (without advancing its lifecycle).",
+    description: "FINDINGS: attach a redacted, SHA-256-tagged evidence record to an existing canonical finding (without advancing its lifecycle). Pass role (proof = the request/response that triggered it; baseline = the normal request showing the differential; verification = the negative control/marker; supporting = anything else) so the finding's evidence stays structured. A reproducible PoC needs proof + baseline + verification.",
     inputSchema: {
       finding: string2().describe("JSON of the Finding object"),
       type: string2().describe("Evidence type (source_location/response/validation_result/negative_control/...)"),
       description: string2().describe("Evidence description"),
+      role: _enum(["proof", "baseline", "verification", "supporting"]).optional().describe("Evidence role (default supporting)"),
       content: string2().optional().describe("Evidence artifact content")
     }
-  }, async ({ finding, type, description, content }) => {
+  }, async ({ finding, type, description, role, content }) => {
     let f;
     try {
       f = JSON.parse(finding);
     } catch {
       return { content: [{ type: "text", text: JSON.stringify({ error: "invalid finding JSON" }) }] };
     }
-    const ev = makeEvidence({ type, description, artifacts: content ? [{ name: "artifact", kind: "evidence", content }] : [] });
+    const ev = makeEvidence({ type, role, description, artifacts: content ? [{ name: "artifact", kind: "evidence", content }] : [] });
     const updated = attachEvidence(f, ev);
-    return { content: [{ type: "text", text: JSON.stringify(updated) }] };
+    return { content: [{ type: "text", text: JSON.stringify({ ...updated, roles: evidenceRoles(updated.evidence) }) }] };
+  });
+  server.registerTool("finding_evidence_roles", {
+    title: "Summarize a finding's evidence by role",
+    description: "FINDINGS: given a finding, return its evidence broken down by role (proof/baseline/verification/supporting) + a `reproducible` flag (true only when it has proof + baseline + verification). Check this before submitting a report: a finding that is not `reproducible` is missing the trigger/differential/negative-control evidence a HackerOne-grade PoC needs.",
+    inputSchema: {
+      finding: string2().describe("JSON of the Finding object")
+    }
+  }, async ({ finding }) => {
+    let f;
+    try {
+      f = JSON.parse(finding);
+    } catch {
+      return { content: [{ type: "text", text: JSON.stringify({ error: "invalid finding JSON" }) }] };
+    }
+    const roles = evidenceRoles(f.evidence ?? []);
+    const missing = [
+      !roles.proof ? "proof (trigger request/response)" : null,
+      !roles.baseline ? "baseline (normal request)" : null,
+      !roles.verification ? "verification (negative control/marker)" : null
+    ].filter(Boolean);
+    return { content: [{ type: "text", text: JSON.stringify({ ...roles, missing: missing.length ? missing : undefined, hint: roles.reproducible ? "reproducible — evidence complete" : "NOT reproducible — attach the missing roles via finding_attach_evidence" }) }] };
+  });
+  server.registerTool("retest_context", {
+    title: "Build the retest context for a confirmed finding",
+    description: "RETEST: given a CONFIRMED (or still_vulnerable/inconclusive) finding, return everything needed to re-verify it in one shot: the finding's key fields (target/method/param/severity), its PoC evidence grouped by role (proof/baseline/verification) + reproducible flag, the active operation constraints (red lines), and the current goal state. Then re-run the proof request and record the verdict with retest_finding.",
+    inputSchema: {
+      finding: string2().describe("JSON of the Finding object to retest")
+    }
+  }, async ({ finding }) => {
+    let f;
+    try {
+      f = JSON.parse(finding);
+    } catch {
+      return { content: [{ type: "text", text: JSON.stringify({ error: "invalid finding JSON" }) }] };
+    }
+    const ev = f.evidence ?? [];
+    const roles = evidenceRoles(ev);
+    const byRole = { proof: [], baseline: [], verification: [], supporting: [] };
+    for (const e of ev)
+      byRole[e.role].push({ evidence_id: e.evidence_id, type: e.type, description: e.description, artifacts: e.artifacts.map((a) => a.name) });
+    const missing = ["proof", "baseline", "verification"].filter((r) => !roles[r]);
+    const goals = goalStatus();
+    return {
+      content: [{
+        type: "text",
+        text: JSON.stringify({
+          finding: {
+            id: f.id,
+            status: f.status,
+            title: f.title,
+            severity: f.classification?.severity,
+            cwe: f.classification?.cwe,
+            target: f.target,
+            method: f.target?.method,
+            param: f.target?.param
+          },
+          poc: { ...byRole, reproducible: roles.reproducible, missing: missing.length ? missing : undefined },
+          constraints: listConstraints(),
+          goals: { all_met: goals.all_met, met: goals.met, total: goals.total },
+          retest_hint: "Re-run the proof request (or the original PoC path) against the CURRENT target. Still reproduces → retest_finding(verdict=still_vulnerable); no longer reproduces → fixed; a blocker (WAF/target down/missing login) prevents a verdict → inconclusive."
+        })
+      }]
+    };
   });
   server.registerTool("redact", {
     title: "Redact secrets from text",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "blitzstrike",
-  "version": "2.4.174",
+  "version": "2.4.176",
   "description": "Blitz Strike — a universal MCP security-audit toolbelt. BLITZ sweeps the attack surface, EAGLE-EYE traces source-to-sink, STRIKE verifies live. 57 attack chains, 130-tool catalog, intelligence data layer. One server, every agent.",
   "type": "module",
   "bin": {