blitzstrike 1.0.66 → 1.0.69

package/CHANGELOG.md

@@ -5,11 +5,28 @@ 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/).
 
+## [1.0.69] — 2026-09-13
+
+### Changed — agent-native skills polish
+
+- `bs-source-audit` now points the driving agent at the dedicated `route_scan`
+  and `complex_scan` detectors for the deep/complex classes.
+- `bs-orchestrate-engagement` gained a skill-chain map (which skill to load per
+  phase) and a stale-server pitfall (reload MCP if a detector returns
+  suspiciously empty output after an upgrade).
+
+## [1.0.67] — 2026-09-13
+
+### Changed — internal cleanup
+
+- Removed stale internal references from source comments and docs so public
+  artifacts are self-contained and generic.
+
 ## [1.0.66] — 2026-09-13
 
 ### Changed — LLM-driven architecture (Mission Control removed)
 
-Aligned with the OMOP model: the LLM is the brain, Blitz Strike is the
+Aligned with the LLM-driven model: the LLM is the brain, Blitz Strike is the
 deterministic hands. The deterministic "self-driving" engine is gone.
 
 - **Removed** `src/mission/` (Mission Control loop, 13 deterministic agents,
@@ -363,8 +380,8 @@ STRIKE core without replacing it.
 
 ### Added
 
-- **Autonomous no-babysitting orchestration** (the missing piece vs.
-  oh-my-open-pentest's plugin loop):
+- **Autonomous no-babysitting orchestration** (the missing piece vs. a
+  harness-level plugin loop):
   - `runEngagement` now returns the full deliverable **inline** — summary,
     markdown, JSON, and SARIF — so a single call completes a source audit
     (no separate `report` call needed).

package/docs/ARCHITECTURE-AUDIT.md

@@ -1,6 +1,6 @@
 # Architecture Audit — Mission Control → LLM-Driven Toolbelt
 
-Step 1 of the OMOP-alignment refactor. Every Mission Control component is
+Step 1 of the LLM-driven refactor. Every Mission Control component is
 classified into one of:
 
 - **TOOL** — a deterministic check the LLM calls via MCP (produces a result the

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "blitzstrike",
-  "version": "1.0.66",
+  "version": "1.0.69",
   "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": {

package/skills/bs-orchestrate-engagement.md

@@ -31,6 +31,17 @@ deterministic MCP tools; never guess a finding — produce it from tool output.
 
 6. **Report.** Emit `generate_report` + `cvss_score` + `write-report`.
 
+## The skill chain (load the right one per phase)
+
+| Phase | Skill |
+|---|---|
+| Before any live action | `scope-gate` |
+| Full engagement / orchestration | `orchestrate-engagement` (this) |
+| Static source analysis | `source-audit` |
+| Prove a hypothesis is real | `verify-finding` |
+| Anti-hallucination gate | `adversarial-review` |
+| Final deliverable | `write-report` |
+
 ## Rules
 
 - One hypothesis per finding; `finding_create` + `finding_transition` for the
@@ -40,3 +51,7 @@ deterministic MCP tools; never guess a finding — produce it from tool output.
   before re-discovering the same target/pattern.
 - You are the orchestrator — do not try to reimplement planning in prose; pick
   tools, act, and let tool output drive the next step.
+- If a detector returns suspiciously empty output for a file that clearly
+  contains a sink, the MCP server may be stale (running old code after a
+  Blitz Strike upgrade). Reload it (`/reload-mcp`) before concluding "no
+  findings" — a stale engine is a silent false negative, not a clean result.

package/skills/bs-source-audit.md

@@ -24,13 +24,13 @@ hypothesis. Do not hand-guess; let the tools produce the source→sink evidence.
    (SQLi → RCE, upload → shell, LFI → RCE). `list_chains` to see the full
    escalation taxonomy. `tech_correlation` for framework-specific weaknesses.
 
-4. **Deep/complex classes.** Beyond plain taint, look for:
-   - **Route confusion** — `call_user_func`, `$obj->$method()`, batch/proxy
-     forwarding, dynamic include, route match without normalization.
-   - **Deserialization → POP chain** — `unserialize` + `__destruct`/`__wakeup`/
-     `__toString` present.
-   - **Type juggling** — loose `==`/`!=` of user input vs hash/secret.
-   - **Mass assignment** — `extract()` without EXTR_SKIP, `parse_str()` of input.
+4. **Deep/complex classes.** Beyond plain taint, drive the dedicated detectors:
+   - `route_scan <file>` — route confusion / dispatch abuse (`call_user_func`,
+     `$obj->$method()`, batch/proxy forwarding, dynamic include, route match
+     without path normalization).
+   - `complex_scan <file>` — deserialization→POP chain (`unserialize` +
+     `__destruct`/`__wakeup`/`__toString`), type juggling (loose `==`/`!=` vs
+     hash/secret), mass assignment (`extract()`/`parse_str()` without a safe flag).
 
 5. **Correlate.** `cve_correlation` + `nvd_lookup` for the framework/pattern to
    confirm the weakness class is real and known.