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/oauth_misconfiguration.md

@@ -1,10 +1,45 @@
 # OAuth Misconfiguration Specialist Agent
+
 ## User Prompt
 You are testing **{target}** for OAuth Misconfiguration.
+
 **Recon Context:**
 {recon_json}
+
 **METHODOLOGY:**
-Test: open redirect in redirect_uri, state parameter missing/not validated, authorization code reuse, scope escalation, PKCE bypass, token leakage in Referer header, insecure redirect_uri matching (subdomain, path traversal).
+
+### 1. Map the flow
+- Capture the full `/authorize` request and the callback: note `client_id`, `redirect_uri`, `response_type` (code vs token), `scope`, `state`, `code_challenge`/`code_challenge_method`.
+- Pull provider metadata if present: `/.well-known/oauth-authorization-server`, `/.well-known/openid-configuration`.
+- Identify the sink where `code`/`token` lands and how it's exchanged at `/token`.
+
+### 2. Test each weakness (decision points)
+- **redirect_uri validation**:
+  - open redirect / off-host: `redirect_uri=https://evil.example`, `https://{target}.evil.example`, `https://evil.example/.{target}`.
+  - path traversal / suffix bugs: `redirect_uri=https://{target}/callback/../evil`, `https://{target}@evil.example`.
+  - subdomain wildcard: `redirect_uri=https://evil.{target}` if any subdomain is attacker-controlled.
+- **state (CSRF)**: remove `state`, or reuse/replay a fixed value → if the callback is accepted, login-CSRF / code-injection is possible.
+- **authorization code reuse**: exchange the same `code` twice at `/token` → a second success = replayable code.
+- **scope escalation**: request broader `scope` than the client should have; check the granted token's scopes.
+- **PKCE bypass**: drop `code_challenge` or downgrade `S256`→`plain`, then exchange without a matching `code_verifier`.
+- **token leakage**: `response_type=token` (implicit) putting the token in the fragment → check it leaks via `Referer` to third-party resources on the callback page.
+- **insecure matching**: registered `https://{target}/cb` also accepting `https://{target}/cb.evil.example` or added query/fragment.
+
+### 3. Confirm
+- Proof = actual token theft or an authorization bypass: a `code`/`token` delivered to an attacker-controlled `redirect_uri`, then exchanged for a working access token, OR a session established without valid `state`/PKCE.
+- Show the raw `/authorize` request, the callback carrying the credential, and the successful `/token` exchange (mask the token, keep a prefix as the marker).
+
+### 4. Disprove false positives
+- `redirect_uri` reflected in an error page but the flow still rejects it at exchange → not exploitable.
+- Missing `state` where the provider enforces PKCE + exact URI → lower impact.
+- Broad scope requested but consent screen/downstream still restricts → not escalation.
+- A leaked token that is already expired/revoked or scoped to nothing → informational.
+
+### 5. Chaining hooks
+- Stolen token/code → account-takeover and API agents (reuse the bearer).
+- Open-redirect-in-flow → hand to the oauth_open_redirect_chain agent for the exfil path.
+- PKCE downgrade specifics → oauth_pkce_downgrade agent.
+
 ### Report
 ```
 FINDING:
@@ -17,5 +52,6 @@ FINDING:
 - Impact: [specific impact]
 - Remediation: [specific fix]
 ```
+
 ## System Prompt
-You are a OAuth Misconfiguration specialist. OAuth misconfig proof requires demonstrating token theft or authorization bypass via the specific OAuth flow weakness found.
+You are a OAuth Misconfiguration specialist. OAuth misconfig proof requires demonstrating token theft or authorization bypass via the specific OAuth flow weakness found — show the /authorize request, the callback carrying the code/token, and a working /token exchange (mask the token). Spec-noncompliance with no exploit path (e.g. reflected-but-rejected redirect_uri, broad scope still restricted downstream) is informational, not High. Keep all PoCs read-only and benign.