Skip to content
KitploitKITPLOIT
ToolsExploitsBlog
Log in
Submit
ToolsExploitsBlog
Submit

Hacking, PenTest, and Cybersecurity Tools for Your Security Arsenal!

Kitploit is a directory of hacking, cybersecurity, and pentesting tools. Discover the latest project updates to find vulnerabilities, analyze systems, automate testing, and strengthen your security.

FeedsContactPrivacy© 2026 Kitploit

Tool Directory

Categories

View all categories
Loading categories
Tools/GitHubGitHub/basesixtythree/b63-poc-cve-2026-63077
Vulnerability AnalysisExploitationWeb Application ExploitationPenetration TestingRemote Access ToolPayload DevelopmentLabs & Practice
GitHub
basesixtythree/b63-poc-cve-2026-63077

B63-POC-CVE-2026-63077

PoC and verification material for CVE-2026-63077 (JetBrains TeamCity agent-polling deserialization RCE)

View Repository
28 days agoNot yet reviewed

Most Popular

View all →

Discover the most used tools by our community.

  • 1

    OsintHunter

    letox4/osinthunter

    🎯 The ultimate OSINT toolkit: investigation graph, confidence scoring, alias families, map, timeline and reports. 49 modules, 400+ sites, no API keys required.

    938
  • 2

    ghostlock-app

    yukonga/ghostlock-app

    GhostLock One-Tap Execution App (CVE-2026-43499)

    556
  • 3

    redforge

    cybersecurity91/redforge

    Browser-side red team toolkit with 102 offline tools for recon, payload generation, credential attacks, crypto, encoding and engagement reporting.

    352
  • 4

    ARTEX

    baarai/artex

    LLM multi-agent autonomous penetration testing platform with planner/worker agents, asset and exploration graphs, MITM traffic recording, and vulnerability retesting.

    343
  • 5

    looking-glass

    anonghosty/looking-glass

    CSIRT platform that ingests Shadowserver reports into MongoDB and replays them as interactive maps, analysis, findings, advisories and executive dashboards.

    336
  • 6

    CyberStrikeAI

    ed1s0nz/cyberstrikeai

    The system of action for AI-native cybersecurity—where intent becomes governed execution, evidence becomes operational memory, and every operation improves the next.

    291

Explore all tools

Browse our collection of tools

View all tools →
Share

CVE-2026-63077 — JetBrains TeamCity agent-polling deserialization RCE

PoC and verification material for CVE-2026-63077 (GHSA-94gx-v738-fx9w, "unauthenticated remote code execution in JetBrains TeamCity", CVSS v3.1 9.8 Critical, CWE-502) — fixed in TeamCity 2026.1.3 and 2025.11.7.

Overview — TeamCity's agent polling protocol is reachable without any credentials. POST /app/agents/v1/register issues a TeamCity-AgentSessionId to anyone, and /app/agents/v1/commands/error deserialises the request body with an XStream mapper whose type allowlist is not exclusive (the mapper never installs NoTypePermission.NONE first, so XStream's default permissions stay in force). An attacker-reachable HTTP port is therefore enough to submit an XML object graph that TeamCity rebuilds into live server objects and ends in a JDBC getConnection() call — bent into an arbitrary file write through HSQLDB's SCRIPT statement, where the written row is itself a JSP scriptlet: a working web shell and command execution as the server user. The fix installs NoTypePermission.NONE before the allowlist, making it deny-by-default.

Verification status (tier-separated)

TierStatusEvidence
Reliability: 10/10 runs per exploit variant on 2026.1.2 (uid=1000(tcuser) returned each run)capturedevidence/01-reliability-readable-10x.txt, evidence/02-reliability-opt-10x.txt
Patched-build negative on 2026.1.3, efficiency variant: FAILED, no webshell, no outputcapturedevidence/03-negative-patched-opt.txt
Patched-build negative on 2026.1.3, readable variant: same rejectioncapturedevidence/04-negative-patched-readable.txt
Max-impact demonstration: unauthenticated RCE as tcuser — id/host/kernel + exfiltration of the server's database connection settings from the data directorycapturedevidence/05-impact-secret-exfiltration.txt
Server-side IOC pair: vulnerable build logs the XStream ConversionException executing the gadget, patched build logs ForbiddenClassException blocking itcapturedevidence/06-server-side-ioc.txt
Per-run timing across both variants (corrected millisecond measurement)capturedevidence/07-timing-comparison.txt

Evidence transcripts are kept as captured; the leading Command:/CWD: headers had the operator home path masked as ~ (visible mask, nothing else touched). Some Command: headers reference the foundry's internal workspace layout (~/.b63/projects/...) and internal driver filenames; the shipped poc/ drivers reproduce the same chain, and lab/ reproduces the captured environment.

Files

poc/exploit_opt.py            compact driver: register -> error -> webshell GET, one keep-alive connection
poc/exploit_readable.py       same chain, stage-by-stage tutorial code
lab/compose.yml               official vendor images pinned by digest: 2026.1.2 (8111) vs 2026.1.3 (8112)
lab/teamcity-vuln/config/database.properties    pre-seeded internal-HSQLDB data dir, vulnerable twin
lab/teamcity-patched/config/database.properties pre-seeded internal-HSQLDB data dir, patched twin
lab/confirm_teamcity_start.py walks TeamCity's first-start wizard screens unattended (no admin setup)
evidence/                     captured lab transcripts (paths scrubbed, see above)

Usage (lab targets only)

Both PoCs speak to local lab instances; the same chain works against any TeamCity instance you are explicitly authorized to test. Python 3.10+ standard library only.

# one command, output returned
python3 poc/exploit_opt.py --url http://127.0.0.1:8111 --cmd "id"

# interactive shell
python3 poc/exploit_opt.py --url http://127.0.0.1:8111
b63> cat /data/teamcity_server/datadir/config/database.properties
b63> exit

Knobs (both variants): --url (default http://127.0.0.1:8111), --cmd (run one command and exit; omit for a shell), --webroot (the only environment-dependent knob — default ../webapps/ROOT, right for the official Docker image whose JVM cwd is /opt/teamcity/bin; WAR deployments need the external container's web application root, absolute paths accepted), --timeout (per-request, default 30 s).

Expected on a vulnerable target: two [b63] stage markers (session issued, webshell dropped) then raw command output, e.g. uid=1000(tcuser) gid=1000(tcuser) groups=1000(tcuser). The command channel is persistent — every later GET with a different ?c= command executes again. On the patched build: [b63] FAILED ... and exit 1; nothing is faked. Exit codes: 0 executed, 1 JSP not reachable / patched / target down.

The gadget chain

The payload is a linked-hash-map with three ordered entries:

  1. Allocate + configure a datasource — HSQLMetadataStorage$SchemaMismatchException is a Throwable, hence permitted by XStream's defaults; its declared fields let the graph build an Apache Commons DBCP2 BasicDataSource with a jdbc:hsqldb:mem: URL and three connectionInitSqls (CREATE TABLE ..., INSERT INTO ... VALUES ('<% JSP scriptlet %>'), SCRIPT '<webroot>/<rand>.jspws').
  2. Expose it to FreeMarker — freemarker.ext.beans.HashAdapter is a Map, hence permitted; its wrapper's inherited object field is re-pointed via an XPath reference at the datasource allocated in entry 1.
  3. Force the getter — a TiedMapEntry (a Map.Entry) inside a HashSet (a Collection): HashSet.add → hashCode → getValue → map.get("connection") → FreeMarker bean property read → BasicDataSource.getConnection() → DBCP runs the init SQL.

HSQLDB's SCRIPT statement dumps the in-memory database to an attacker-chosen path relative to the server's working directory; the dumped row contains the JSP scriptlet verbatim, so the SQL dump is a JSP page (an SQL/JSP polyglot — the same trick Secfault Security used against LibreOffice). TeamCity maps .jspws to Jasper; the first anonymous GET compiles and executes it.

Wire summary: POST /app/agents/v1/register → session header; POST /app/agents/v1/commands/error → HTTP 500 after the SQL has run (500 is the success case); GET /<rand>.jspws?c=<command> → output between per-run random markers.

Note on reuse: earlier public research on this CVE ships a one-shot PoC — it proves the bug with a single Runtime.exec() and then deletes its own JSP — no output, no reuse. The shipped PoCs keep the gadget chain intact and replace the impact stage with a ProcessBuilder shell returning stdout/stderr, so each request is a full command round trip.

Lab reproduce

cd lab
docker compose up -d                   # official images pinned by digest, loopback-only ports
sleep 60
V=$(docker logs cve-2026-63077-vuln 2>&1 | grep -oE 'Super user authentication token: [0-9]+' | head -1 | grep -oE '[0-9]+')
P=$(docker logs cve-2026-63077-patched 2>&1 | grep -oE 'Super user authentication token: [0-9]+' | head -1 | grep -oE '[0-9]+')
python3 confirm_teamcity_start.py --super-token "$V" --super-token "$P"
docker compose ps
Download Tool