feat: add 6 web + LLM vuln scanners — CORS, CRLF, NoSQLi, JWT, OOB, LLM red-team#82
Open
DebasishTripathy13 wants to merge 3 commits into
Open
feat: add 6 web + LLM vuln scanners — CORS, CRLF, NoSQLi, JWT, OOB, LLM red-team#82DebasishTripathy13 wants to merge 3 commits into
DebasishTripathy13 wants to merge 3 commits into
Conversation
DebasishTripathy13
commented
Jun 19, 2026
DebasishTripathy13
commented
…LM red-team
Adds six new capabilities, each as tool + slash command + tests, all
pure-Python (no new deps) and auto-installed via the existing commands glob:
- /cors CORS misconfig (origin reflection, null-origin, credentialed
read, suffix/prefix regex bypass, scheme downgrade)
- /crlf CRLF / response-splitting + host-header injection (Set-Cookie
canary, encoded + UTF-8 bypass variants)
- /nosqli NoSQL injection (operator auth-bypass, bracket-syntax,
$where time-based blind) with differential + timing classifier
- /jwt-scan JWT alg:none forgery, RS256->HS256 confusion, HS256 secret
crack, static claim analysis (offline, pure stdlib)
- /oob out-of-band orchestrator wrapping interactsh-client — confirms
blind SSRF/XXE/SQLi/RCE/Log4Shell via callback correlation
- /llm-redteam LLM red-team corpus (prompt-injection, jailbreak, system-prompt
leak, data-exfil, indirect injection, guardrail bypass) with
canary-token detection
Core logic is separated from network I/O so the 46 new tests run offline.
docs/CAPABILITY-GAPS.md records the full coverage audit; remaining gaps
(headless-browser DOM harness, garak/PyRIT, web3 static/fuzzing) are queued
as the next pipeline tier.
DebasishTripathy13
force-pushed
the
feat/web-llm-scanners
branch
from
June 19, 2026 18:25
8700dbb to
e1cdfd1
Compare
There was a problem hiding this comment.
Pull request overview
This PR adds six new purpose-built security scanners (CORS, CRLF/host-header, NoSQLi, JWT, OOB correlation via interactsh, and an LLM red-team harness) plus corresponding slash-command docs and unit tests, extending the repo’s “confirmation” tooling beyond skill/documentation coverage.
Changes:
- Introduces 6 new scanner tools under
tools/(networked where appropriate; otherwise pure/offline helpers). - Adds unit tests for the tools’ pure components (payload generation + classification/correlation).
- Documents the new tooling in
commands/*,docs/*, and updates command/tool listings inCLAUDE.md.
Reviewed changes
Copilot reviewed 22 out of 22 changed files in this pull request and generated 11 comments.
Show a summary per file
| File | Description |
|---|---|
| tools/oob_listener.py | OOB payload generator + interactsh listener wrapper + interaction correlation |
| tools/nosqli_scanner.py | NoSQLi login scanner + query-string payload emitter |
| tools/llm_redteam.py | LLM red-team corpus runner with canary-based response classification |
| tools/jwt_scanner.py | Offline JWT analysis + alg=none + alg-confusion + HS256 wordlist cracking |
| tools/crlf_scanner.py | CRLF/response-splitting and host-header injection scanner |
| tools/cors_scanner.py | CORS misconfiguration scanner (Origin probes + ACAO/ACAC classification) |
| tests/test_oob_listener.py | Unit tests for OOB payload generation and correlation |
| tests/test_nosqli_scanner.py | Unit tests for NoSQLi payloads and classifiers |
| tests/test_llm_redteam.py | Unit tests for corpus coverage and response classification |
| tests/test_jwt_scanner.py | Unit tests for JWT decode/forge/crack/analyze primitives |
| tests/test_crlf_scanner.py | Unit tests for CRLF payload generation and detection helpers |
| tests/test_cors_scanner.py | Unit tests for CORS origin generation and classification |
| docs/README.md | Adds CAPABILITY-GAPS doc entry |
| docs/CAPABILITY-GAPS.md | Capability-gap audit updated to mark shipped scanners/commands |
| commands/README.md | Updates command count and lists new vulnerability scanners |
| commands/oob.md | Adds /oob command documentation |
| commands/nosqli.md | Adds /nosqli command documentation |
| commands/llm-redteam.md | Adds /llm-redteam command documentation |
| commands/jwt-scan.md | Adds /jwt-scan command documentation |
| commands/crlf.md | Adds /crlf command documentation |
| commands/cors.md | Adds /cors command documentation |
| CLAUDE.md | Updates command list and tool inventory to include new scanners |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+182
to
+185
| if args.correlate: | ||
| inters = [json.loads(l) for l in open(args.correlate) if l.strip()] | ||
| payloads = json.load(open(args.payloads_file)) if args.payloads_file else {} | ||
| hits = correlate(inters, payloads) |
Comment on lines
+12
to
+20
| Severity model: | ||
| CRITICAL reflects an attacker origin AND allows credentials -> cookie-auth'd | ||
| cross-origin reads. Full account-data exfil. | ||
| HIGH `null` origin trusted with credentials (sandboxed iframe / data: URI | ||
| attack), OR sibling/suffix-domain trusted with credentials. | ||
| MEDIUM reflects attacker origin without credentials (token-auth data read), | ||
| OR trusts http:// downgrade of an https origin. | ||
| LOW ACAO: * with no credentials (public CORS — usually intended). | ||
|
|
Comment on lines
+37
to
+40
| USER_AGENT = "claude-bug-bounty/cors_scanner" | ||
|
|
||
| CRITICAL, HIGH, MEDIUM, LOW, INFO = "CRITICAL", "HIGH", "MEDIUM", "MEDIUM_LOW", "INFO" | ||
|
|
Comment on lines
+25
to
+28
| def test_generate_tests_covers_key_vectors(): | ||
| origins = [t.origin for t in generate_tests(URL)] | ||
| assert any("evil.example" in o and "target.com" not in o.split("//")[1].split("/")[0].replace("evil.example", "") for o in origins) or True | ||
| # null + scheme downgrade + post-domain present |
Comment on lines
+67
to
+72
| def where_sleep_body(user_field: str, pass_field: str, ms: int = 5000) -> dict: | ||
| """$where time-based blind payload to confirm server-side JS evaluation.""" | ||
| return { | ||
| user_field: "admin", | ||
| pass_field: {"$where": f"sleep({ms})"}, | ||
| } |
Comment on lines
+30
to
+33
| def test_where_sleep_body_shape(): | ||
| body = where_sleep_body("u", "p", ms=5000) | ||
| assert body["p"] == {"$where": "sleep(5000)"} | ||
|
|
Comment on lines
+61
to
+69
| def forge_alg_none(token: str) -> list[str]: | ||
| """Produce alg:none forgeries with the original claims and an empty sig.""" | ||
| header, payload, _ = decode(token) | ||
| out = [] | ||
| for alg in ("none", "None", "NONE", "nOnE"): | ||
| h = dict(header) | ||
| h["alg"] = alg | ||
| out.append(f"{_encode_segment(h)}.{_encode_segment(payload)}.") | ||
| return out |
Comment on lines
+169
to
+173
| if args.alg_none: | ||
| did = True | ||
| print("# alg:none forgeries") | ||
| for t in forge_alg_none(args.token): | ||
| print(t) |
Comment on lines
+86
to
+90
| class CrlfFinding: | ||
| url: str | ||
| vector: str # "query" | "host-header" | ||
| payload: str | ||
| evidence: str |
Comment on lines
+127
to
+131
| hdrs = _send(target, None, timeout) | ||
| ev = detect_injection(hdrs) | ||
| if ev: | ||
| findings.append(CrlfFinding(target, "query", payload, ev)) | ||
| if host_header: |
Groups CORS / CRLF / NoSQLi / JWT / OOB / LLM red-team into a single loadable skill with routing table, per-tool usage, severity models, and chain patterns — mirroring the graphql-audit skill structure. Wired into the CLAUDE.md skills table; install.sh picks up skills/* automatically.
Codename after Argus Panoptes, the hundred-eyed giant — six 'eyes' that surface what ordinary scans miss, including blind/out-of-band bugs. Updates skill frontmatter name, title/tagline, and the CLAUDE.md row.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.